Text preview for : Jitter Analysis Using Keysight_2527s InfiniiVision 6000 X-Series and Infiniium Series - Application part of Agilent Jitter Analysis Using Keysight 2527s InfiniiVision 6000 X-Series and Infiniium Series - Application Agilent Jitter Analysis Using Keysight_2527s InfiniiVision 6000 X-Series and Infiniium Series - Application Note 5991-4000EN c20140915 [13].pdf
Back to : Jitter Analysis Using Key | Home
Keysight Technologies
Jitter Analysis
Using InfiniiVision 6000 X-Series and
Infiniium Series Oscilloscopes
Application Note
Introduction
As data rates continue to increase in today's state-of-the-art
high-speed digital designs, timing budgets are decreasing.
Ensuring that serial data signals are valid and stable when
receivers sample the data often requires an understanding of the
effects of the various components of jitter that may contribute
to decreased valid data windows. The primary measurement
tool used today by hardware design engineers to capture and
view waveform jitter is an oscilloscope. Many of today's higher
performance oscilloscopes also provide optional jitter analysis
measurement capabilities that can not only be used to view jitter
in different display formats, but they can also quantize the various
components of jitter.
This application note begins with a discussion of various display
formats that can be used to view jitter, including horizontal
waveform histograms, TIE histograms, TIE trend waveforms,
and jitter spectrum waveforms. Also discussed are various clock
recovery algorithms, real-time eye displays, jitter separation,
and total jitter extrapolation. This application note wraps up
with a comparison of how both the Keysight Technologies, Inc.
6000 X-Series and Infiniium Series oscilloscopes address these
challenges.
Composite views of jitter and clock recovery
Let's begin by defining jitter. Jitter is the deviation of a timing event of a signal
from its ideal position. This is often times referred to as Time Interval Error (TIE).
Engineers often think of jitter as a "bouncing edge" relative to a reference (trigger
point) as shown in Figure 1. In this measurement example, we are repetitively
capturing an edge of a clock signal multiple cycles after a trigger reference edge
on the same signal. One critical factor that enhances the scope's ability to quickly
capture and display extremes of jitter such as this is fast waveform update rate.
Also shown in Figure 1 is a waveform histogram plot of the jitter based on a narrow
slice of waveform data around the 50% amplitude level. A horizontal waveform
histogram plot displays the probability distribution function (PDF) of the composite
jitter time-correlated to the captured waveforms. In this example we can see that
the jitter has a bi-modal distribution. Waveform histogram measurements, which
are the simplest type of jitter measurements and are one of the key measurement
capabilities found in most of today's higher performance scopes, including
Keysight's InfiniiVision 6000 X-Series and Infiniium Series oscilloscopes.
Although capturing and viewing jitter on a repetitive clock signal is relatively
straight forward, capturing and viewing jitter on a serial data signal is a bit more
complex, especially if an explicit clock signal for triggering on is not available. Most
of today's higher speed serial busses have embedded clocks (non-explicit) that
must be recovered by receivers. Performing jitter analysis on serial bus data signals
with embedded clocks means that the scope must also be able to recover the clock
from the data signal. Clock recovery in the oscilloscope is produced by a software
algorithm that is part of the jitter analysis option to create a virtual clock that
emulates the clock recovery of a serial data bus receiver. This virtual clock is then
used as a reference waveform for which to perform multiple TIE measurements
on consecutive data edges based on a deep memory acquisition of the serial bus
signal.
Figure 1: Basic Jitter analysis; viewing repetitive clock jitter with a waveform histogram plot. The histogram (blue)
shows that the jitter of the clock (yellow) has bi-modal distribution.
3
Figure 2 shows an example of repetitive acquisitions of a continuous serial data
signal. Since we don't have a reference signal -- other than the signal itself -- to
use as a trigger source, we can't view the timing error of each data edge (jitter). To
perform jitter analysis on this waveform, we will need to use the scope's jitter analysis
option with clock recovery.
Figure 2: Repetitively capturing a continuous serial bus data signal while triggering on the data signal.
Using the scope's jitter analysis option, we first set up the scope to capture multiple
bit periods of the data signal using the scope's deep memory acquisition. A bit period
of a serial data signal is typically called a Unit Interval (UI). We then perform jitter
analysis on Data TIE based on a constant clock recovery algorithm.
Figure 3 shows the measured jitter in a histogram plot. However, the data that
produced this histogram plot is very different than the data used to produce the
histogram plot in the previous measurement example (Figure 1) of clock jitter. Rather
than producing the histogram plot
from captured waveform data relative
to a trigger point, this plot is based on
measured timing error (TIE) of every Outliers
captured data edge (rising and falling)
relative to the recovered clock (not
shown). In addition to displaying the
histogram, today's higher performance
oscilloscopes also show full statistics of
the measurement histogram (refer to far
right panel of the scope's display in Figure
3).
From this histogram plot we can see that
the jitter in this serial data signal appears
to have a strong Gaussian (random)
component as well as a bi-modal
deterministic component. Also notice
that the histogram plot shows extreme
"outliers" near the lower-right side of the
scope's waveform display area. Let's now
view this timing error in another display/
plot format that may reveal additional
information about this jitter. Figure 3: Jitter analysis using a TIE histogram plot.
4
Figure 4 shows the same jitter
analysis with the addition of the TIE Outlier
jitter trend waveform display (purple
trace). The TIE jitter trend waveform
is a plot of timing error of each data
edge relative to the recovered clock
(vertical axis) versus time (horizontal
axis). This waveform is time-correlated
to the captured serial data signal
(yellow trace). From the TIE jitter trend
plot we can see that our data signal
has sinusoidal modulation, which
correlates to the bi-modal distribution
of the jitter histogram plot. In addition
to the sinusoidal modulation, it
appears that our data signal has
occasional extreme positive timing
errors as evidenced by the spikes in
the TIE jitter trend waveform. This is
what is producing the "outliers" in the
histogram plot.
Figure 4: TIE jitter trend waveform shows sinusoidal modulation along with extreme timing errors.
Determining the frequency of the
modulation might provide us with
a clue as to the source of the
jitter. Some higher performance
oscilloscopes, including the Keysight
Infiniium Series allow you to directly
measure the frequency of the
measurement trend waveform. If this
is not available, one simple method
of measuring the frequency would be
to use the scope's timing cursors to
measure the period of the sinusoidal
modulation. Another option would be
to turn on the jitter spectrum view
(lower purple trace) as shown in
Figure 5. The jitter spectrum waveform Data TIE spectrum
is produced by running an FFT math
operation on the TIE trend waveform
(upper purple trace). All three of
these methods allow us to find the
frequency of modulation, which in this
case is at exactly 20 kHz. Figure 5: Using the jitter spectrum view to display and measure the frequency of jitter modulation. This signal has a
frequency of 20 kHz giving us some insight into where that jitter is coming from.
In the above jitter measurement examples, the clock recovery algorithm
automatically found the nominal bit rate of the serial bus signal based on a
constant clock. Also available are 1st order and 2nd order Phase Lock Loop (PLL)
clock recovery algorithms that can be selected. Many of today's high-speed digital
systems actually modulate their transmitted data intentionally. This is referred to as
spread spectrum clocking (SSC), which is primarily used to reduce Electromagnetic
Interference (EMI). If this is case, then a PLL type clock recovery algorithm should
be selected to emulate the PLL clock recovery of receivers in SSC systems.
5
Figure 6 shows jitter measurements on the same serial bus data, but this time
using a 1st order PLL clock recovery algorithm with a specified loop bandwidth
of 200 kHz. Compare Figure 4 with Figure 6. Notice that the 20 kHz sinusoidal
modulation has been virtually eliminated from the TIE trend waveform in Figure 6,
and the jitter histogram is beginning to look less bi-modal and more Gaussian.
Outliers
Figure 3: Jitter analysis using a TIE histogram plot.
Outlier
Figure 6: Jitter measurements using a 1st order PLL clock recovery algorithm makes the jitter histogram when
compared to figure 4 look less bi-modal and more Gaussian.
Figure 4: TIE jitter trend waveform shows sinusoidal
modulation along with extreme timing errors.
Another very insightful view of jitter
within serial data signals is to display
the captured waveform data as a Real
Time Eye (RTE) as shown in Figure
7. A real time eye is an overlay of
every captured UI. Based on the clock
recovery, the scope slices the captured
waveform into individual bit periods
(UIs), and then overlays every slice into
a single color-graded display. From this
display we can visually see a composite
view of peak-to-peak edge jitter and
noise. We also observe extreme
positive shifts in the data edges (dark
blue traces), which is the cause of the
"outliers" revealed in the histogram
plot (Figure 3) and the spikes in the
TIE trend waveform (Figure 4). Also
available is the ability to automatically
measure the eye height and eye width,
which determines the data valid
window.
Figure 7: From this display we see a composite view of peak-to-peak edge jitter and noise with extreme positive shifts
in the data edges (dark blue traces), which is the cause of the "outliers" revealed in the histogram plot and shows the
worst-case eye opening.
6
All the above jitter and real-time eye measurements were performed using a
Keysight 6000 X-Series 6 GHz bandwidth oscilloscope with the DSOX6JITTER
option. These same jitter measurements can also be performed using a Keysight
Infiniium Series oscilloscope with the EZJIT basic jitter analysis option as shown
in Figure 8. Notice that the oscilloscope places the waveform, measurement
histogram, measurement trend, and jitter spectrum all in separate grids. Each
grid can then be used to dive deep into the source of the jitter. The EZJIT option
provides an easy-to-use wizard to set up all of your jitter measurements.
Figure 8: Jitter analysis using a Keysight Infiniium Series oscilloscope with the EZJIT option.
Components of jitter and jitter separation
As we saw in these measurement examples, Total Jitter (TJ) is often composed
of a combination of random and deterministic jitter components. Random Jitter
(RJ) is theoretically unbounded and is usually measured in terms of an RMS
value. "Unbounded" simply means that if you wait long enough and continue to
collect TIE measurement data, the peak-to-peak jitter will increase indefinitely --
theoretically. But the RMS value of RJ will eventually resolve to a stable value. In
addition, random jitter is assumed to be very predictable in terms of distribution.
The Probability Distribution Function (PDF) is always Gaussian in distribution (bell-
shaped curve). Many of today's high-speed standards require measurements of not
only TJ, but also RJ and DJ to show that your device is compliant. TJ
On the other hand, Deterministic Jitter (DJ) is bounded and is always measured Bounded (p-p) Unbounded (RMS)
in terms of a peak-to-peak value. Although the distribution of deterministic jitter DJ RJ
can be very unpredictable, the likely causes and characteristics of the individual
sub-components of measured deterministic jitter are very predictable. The sub-
components of DJ consists of Periodic Jitter (PJ) and Data Dependent Jitter (DDJ),
which includes Duty Cycle Distortion (DCD) and Inter-symbol Interference (ISI)
PJ DDJ
as shown in the Figure 9 illustration. Note that the 20 kHz sinusoidal modulation
shown in the previous measurements is an example of Periodic Jitter. To learn more
about the various components of jitter, refer to the application note titled, "Finding DCD ISI
Sources of Jitter with Real-time Jitter Analysis", listed at the end of this document.
Figure 9: Jitter components.
7
Since TJ is composed of an RJ component, which is theoretically unbounded, TJ
is also unbounded. Directly measuring TJ means that you would need to collected
data indefinitely. For this reason, TJ is always based on a user-defined Bit Error
Ratio (BER) factor, which is typically very low, such as 10 -12. TJ is typically reported
in terms of a peak-to-peak value. But even when based on a low BER factor, it is
still impossible to directly measure TJ with an oscilloscope. TJ must be extrapolated
from the collection of partial data.
Although we were able to determine in the previous measurement examples
that the measured jitter consisted of a combination of jitter components (random
and deterministic), we were only able to measure the composite jitter while
using a Keysight InfiniiVision 6000 X-Series oscilloscope. Separating out the
various components of jitter and extrapolating TJ based on a BER factor requires
more advanced jitter analysis, which is available in Keysight's Infiniium Series
oscilloscopes with the EZJIT+ option.
Figure 10 shows an example of jitter separation using a Keysight Infiniium
oscilloscope. Using a Tail Fit algorithm this oscilloscope with the EZJIT+ option is
able to first extract DJ from RJ,PJ, and then extract RJ from the tail of the RJ,PJ
histogram. It then extrapolates TJ based on a defined BER Bathtub curve. To
learn more about jitter separation techniques, refer to the application note titled,
"Analyzing Jitter using Keysight's EZJIT Plus Software", listed at the end of this
document.
Figure 10: Jitter separation and TJ extrapolation using an Infiniium Series oscilloscope with the EZJIT+ option.
8
Choosing the right oscilloscope platform for your jitter measurements
So which oscilloscope platform from Keysight best fits your jitter measurement
needs; the InfiniiVision 6000 X-Series or the Infiniium Series? This depends on
your oscilloscope performance requirements (bandwidth, sample rate, and memory
depth), jitter measurement requirements (basic jitter or advanced jitter analysis with
separation), your use-model (debug or analysis), and your budget.
The Keysight InfiniiVision 6000 X-Series oscilloscope provides up to 6 GHz
bandwidth with sampling up to 20 GSa/s and a standard memory depth of 4 M
points. This platform is based on a real-time operating system that has been
optimized for debugging high-speed digital designs with Keysight's fourth-
generation MegaZoom technology. With the fastest waveform update rates in this
class of oscilloscope (up to 450,000 waveforms per second), this oscilloscope can
capture infrequent transients that other scopes miss. The InfiniiVision 6000 X-Series
oscilloscope is also the lowest priced scope in this performance category with
optional jitter analysis (DSOX6JITTER). For the cost-conscience engineer, the 6000
X-Series may provide just enough analysis to get the job done.
The Keysight Infiniium Series oscilloscope, including the S-Series shown below
provides up to 63 GHz bandwidth with sampling up to 160 GSa/s. These scopes also
provide the deepest memory in the oscilloscope industry with up to 2 G points of
acquisition memory (500X more than the 6000 X-Series). This platform is based on a
Windows operating system that has been optimized for advanced waveform analysis
of high-speed digital designs. As discussed earlier, the Infiniium oscilloscope features
both EZJIT and EZJIT+ jitter analysis options. While some of the measurements of the
EZJIT option are similar to the jitter measurements available in the InfiniiVision 6000
X-Series with the DSOX6JITTER option, EZJIT offers the flexibility of looking at up to
17 measurement trends and histograms at once.
You can also analyze jitter on multiple waveforms at once. You can take your
analysis even further with the EZJIT+ advanced jitter analysis option, which can
separate jitter into their various sub-components, as well as extrapolate total jitter
based on a BER factor. EZJIT+ provides numerous charts and tools to truly debug
and analyze any jitter that you have in your system.
The Keysight InfiniiVision 6000 X-Series oscilloscope. The Keysight Infiniium S-Series oscilloscope.
9
The following table summarizes the features of Keysight's various jitter analysis options in InfiniiVision 6000 X-Series and Infiniium oscilloscopes:
InfiniiVision 6000X with Infiniium Series with EZJIT Infiniium Series with EZJIT+
DSOX6JITTER option
Jitter views
Jitter trend
Jitter histogram
Jitter spectrum
Real-time eye With SDA Option 1
With SDA Option1
Clock recovery
Constant clock
1st order PLL clock
2nd order PLL clock
Explicit clock
Clock jitter measurements
Clock TIE
N-period
Period to period
+Width to +width
-Width to