只要测试数据通信IC或测试电信网络，就需要测试抖动。抖动是应该呈现的数字信号沿与实际存在沿之间的差。时钟抖动可导致电和光数据流中的偏差位，引起误码。测量时钟抖动和数据信号就可揭示误码源。
测量和分析抖动可借助三种仪器：误码率(BER)测试仪，抖动分析仪和示波器(数字示波器和取样示波器)。
选用哪种仪器取决于应用，即电或光、数据通信以及位率。因为抖动是误码的主要原因，所以，首先需要测量的是BER。若网络、网络元件、子系统或IC的BER超过可接受的限制，则必须找到误差源。
大多数工程技术人员希望用仪器组合来跟踪抖动问题，先用BER测试仪、然后用抖动分析仪或示波器来隔离误差源。
BER测试仪制造商需要测量其产品的BER，以保证产品符合电信标准。当需要表征数据通信元件和系统时，BER测试对于测试高速串行数据通信设备也是主要的。
BER测试仪发送一个称之为伪随机位序列(PRBS)的预定义数据流到被测系统或器件。然后，取样接收数据流中的每一位，并对照所希望的PRBS图形检查输入位。因此，BER测试仪可以进行严格的BER测量，有些是抖动分析仪或示波器不可能做到的。
尽管BER测试仪可进行精确的BER测量，但是，对于10-12BER(每1012位为1位误差)精度的网络或器件测试需数小时。为了把测试时间从数小时缩短为几分钟，BER测试仪采用“BERT
scan”技术，此技术用统计技术来预测BER。
可以编程BER测试仪在位时间(称之为“单位间隔”或“UI”)的任何点取样输入位。“澡盆”曲线表示BER是取样位置的函数。若BER测试仪检测位周期(0.5UI)中心的位，则抖动引起位误差的概率是小的。若BER测试仪检测位于靠近眼相交点上的位，则将增大获得抖动引起位误差的似然性。
抖动分析仪BER测试仪不能提供有关抖动持性或抖动源的足够信息。抖动分析仪(往往称之为定时时间分析仪或信号完整性分析仪)可以测量任何时钟信号的抖动，并提供故障诊断抖动的信息。抖动分析仪也用抖动特性来预测BER，其所用时间比BER测试仪小很多。
抖动测试仪对于测试高速数据通信总线(如光纤通信，SerialATA, Infiniband,
Rapidio，每个通道的数据率高达3.125Gbits/s)用的器件是有用的。因为抖动分析仪在几秒内可预测BER，所以，对于生产线测试是有用的，很多ATE制造商根据用户要求，把抖动测试仪安置在测试系统中。
抖动分析仪检测信号沿并测量沿之间的时间。在采集定时数据之后，抖动分析仪执行算法，产生直方图、频率曲线、数据的其他直观图像。这些图像展示干扰信号的线索。靠执行直方图和频率曲线的计算，抖动分析仪把整个抖动分离为随机抖动和确定性抖动。
比如一种确定性抖动，它具有一个特殊源。一个干扰信号相位调制基准信号来产生测量信号中的抖动。抖动分析仪可以计算呈现在抖动中的频率(相位1-4)。一旦知道抖动频率，就可隔离抖动源并减轻抖动影响。若干扰信号的频率对应于其他时钟频率，则用增加EMI屏蔽或其他方法把源隔离就可解决问题。
示波器两类示波器证明对于抖动测试和分析是有用的。为了测试通信速度达3.125Gbits/s(在铜线上传输数据，这可能是最高速度)的器件、缆线、子系统或系统，可以用实时取样示波器。它们类似于抖动分析仪，可以测量任何时钟信号的抖动。
为了测量光信号，如OC-192和10Gigabit
Ethernet(9.952Gbits/s)或OC-768(39.808Gbits/s)，就需要50GHz~75GHz带宽的取样示皮器(如Agilent数字通信分析仪或Tek通信信号分析仪)。也可在电数据信号中用这些示波器。
宽带示波器对于测试当今所用的最高位率的抖动是有用的。因为它们的低取样率(150ksamples/s或更低)，所以，它们需要重复信号(如PRBS)来建立眼图，它们从眼图可建立抖动直方图。
示波器制造商在其示波器上提供抖动分析软件。
定时误差图是数据流的有效瞬时相位图。它示出抖动包含周期成分。定时误差图的快速傅里叶变换(第3个图线)定标为1MHz/div，显示抖动的频率。此频率可对应于开关电源的时钟频率或来自系统数据缆线中的交扰。
眼图交叉点的直方图显示分布有2个峰。双峰表明确定性抖动，它来自外部干扰(如开关电源)。另一处抖动——随机抖动遵从高斯分析，不能确定它们的源。
混合仪器最近，某些测试设备制造商已开发出混合仪器。传统的BER测试仪只给出位误差，现在BER测试仪执行某些抖动分析，甚至有的还包含取样示波器。现在抖动分析仪也包含取样示波器，如Warecrest
SIA-3000。这些取样示波器可观察眼图，但它们没有专用取样示波器那样的带宽。现在混合仪器的示波器带宽最高为6GHz。实时和等效时间取样示波器现在提供测量抖动和计算BER的软件。
Whether you test datacom ICs that exchange data with other chips on
a board or you test telecom networks that send data many miles, you
need to measure jitter—the difference between when a digital
signal's edges should occur and when they actually occur. Jitter in
clocks can cause misaligned bits in both electrical and optical
data streams, leading to bit errors. By measuring jitter on clock
and data signals, you can uncover the sources of bit errors.
Three types of instruments can help you measure and analyze jitter:
bit-error-rate (BER) testers, jitter analyzers, and oscilloscopes
(both digitizing scopes and sampling scopes). Figure 1 shows the
bit rate that each instrument can handle.
Figure 1 The data rate you use can determine which type of instrument you
need for measuring jitter.
Which type of test instrument you should use depends on your
application—electrical or optical, datacom or telecom—and bit rate.
Because jitter is a major cause of bit errors, you often need to
first measure BER. If the BER of a network, network component,
subsystem, or IC exceeds acceptable limits, then you must find the
error source.
You most likely will need a combination of instruments to track
down jitter problems. Tommy Cook, R&D section manager at
Agilent Technologies (South Queensferry, Scotland), says that many
engineers start with a BER tester, then move to a jitter analyzer
or oscilloscope to isolate the cause of the errors. (The
application described in "BERT and scope " highlights one such
instance.)
BER testersManufacturers need to measure their products' BER to ensure the
products comply with telecom standards. BER testing is also
essential for testing high-speed serial data communications
equipment when you need to characterize datacom components and
systems.
A BER tester sends a predefined data stream, called a pseudorandom
bit sequence (PRBS), through a system or device under test. It then
samples each bit in a received data stream and checks incoming bits
against the expected PRBS pattern. BER testers, therefore, can give
you an exact BER measurement, something you can't get with jitter
analyzers or scopes. (See "BER measurements reveal network health,"
Ref. 1.)
Figure 2 When looking for jitter and for bit errors, you get the best
performance if you sample bits at the center of the eye.
Although BER testers can produce accurate BER measurements,
achieving that accuracy on a network or device that's designed for
a BER of 10–12 (1 bit error for every 1012 bits) can take hours. To
cut testing time from hours to minutes, BER testers employ a "BERT
scan" technique that uses statistical techniques to predict BER.
Figure 2 shows how the scan works using the familiar eye diagram as
a reference. You can program BER testers to sample incoming bits at
any point in a bit's duration (called a "unit interval" or "UI").
The graph below the eye diagram (often called a "bathtub" curve)
shows BER as a function of the sample location. If a BER tester
checks bits at the center of a bit period (0.5 UI), then the
probability that jitter will cause a bit error should be small. But
if the BER tester checks bits at locations closer to the eye's
crossover points, it will increase its likelihood of finding a bit
error caused by jitter.
Jitter analyzersBER testers, however, may not provide enough information about the
characteristics or sources of jitter. Jitter analyzers (often
called timing-interval analyzers or signal-integrity analyzers) can
measure jitter in any clock signal and provide you with information
that can help you troubleshoot jitter. These instruments also use
jitter characteristics to predict BER in considerably less time
than a BER tester.
You'll find jitter analyzers useful for testing devices used in
high-speed datacom buses such as Fibre Channel, SerialATA,
Infiniband, and RapidIO at data rates up to 3.125 Gbits/s per
channel (Ref. 2). Because jitter analyzers predict BER in just
seconds, they are useful for production line testing, and many ATE
manufacturers will install a jitter analyzer—specified by the
customer—into their test systems.
Figure 3 Jitter analyzers and oscilloscopes can separate jitter into its
components. Here, the jitter’s sinusoidal component is clearly
visible, which can reveal clues as to the source of the jitter.
Courtesy of Anritsu.
Jitter analyzers detect a signal's edges and measure the time
between them. After acquiring timing data, the jitter analyzer runs
algorithms that generate histograms, frequency plots, and other
visual images of the data. These graphs often reveal clues that
lead you to interfering signals. By performing calculations of the
histogram and frequency plots, the jitter analyzer separates total
jitter into its components—random jitter and deterministic jitter.
For tutorials on jitter and its components, you can download papers
listed in "Jitter lessons on the Web ."
Figure 3 shows a type of deterministic jitter, which has a specific
source. An interfering signal phase modulates the reference signal
in the upper trace to produce the jitter in the measured signal in
the lower trace. Jitter analyzers can calculate the frequencies
present in jitter (Phases 1–4). Once you know the jitter frequency,
you can often isolate the jitter source and mitigate its effects.
If the interfering signal's frequency corresponds to another clock
frequency, for example, you may be able to solve the problem by
adding EMI shielding or otherwise isolating the source.
OscilloscopesTwo types of oscilloscopes prove useful for jitter measurement and
analysis. To test devices, cables, subsystems, or systems that
communicate at speeds up to 3.125 Gbits/s (the current highest
speeds possible for transmitting data over copper), you can use a
real-time sampling oscilloscope. Like jitter analyzers, these
scopes can measure jitter in any clock signal, not just those used
in communications. (See "Not just a communications problem " for an
application example.)
To make measurements on optical signals such as OC-192 and 10
Gigabit Ethernet (9.952 Gbits/s) or OC-768 (39.808 Gbits/s), you
need the 50-GHz to 75-GHz bandwidth of a sampling scope, such as
the Agilent digital communications analyzer or the Tek
communications signal analyzer. You can use these scopes on
electrical data signals, too.
The high bandwidth of sampling scopes makes them useful for
measuring jitter at the highest bit rates in use today. Because of
their low sampling rates (150 ksamples/s or less), they require
repetitive signals such as PRBS patterns to build eye diagrams from
which they can build jitter histograms.
Figure 4 Oscilloscopes can display a plot of a time-interval error, which
can take on periodic characteristics. A FFT of the time-interval
error reveals the frequency, and a histogram reveals the jitter
distribution. Courtesy of LeCroy.
Oscilloscope manufacturers offer jitter-analysis software on their
scopes, and Figure 4 shows some of the jitter information you can
extract with a scope. (You can also get this type of analysis with
a jitter analyzer.) The uppermost trace represents an OC-48 (2.488
Gbits/s) PRBS pattern that repeats every 127 bits. In the second
trace, the scope calculates the timing error of each bit against a
software-generated ideal clock.
The timing-error plot is effectively an instantaneous phase plot of
the data stream. It shows that the jitter contains a periodic
component. A fast Fourier transform of the timing-error plot (third
trace, in blue), scaled to 1 MHz/div, reveals the frequency of the
jitter. That frequency could correspond to a switching power
supply's clock frequency or it could come from crosstalk in the
system's data cables.
Figure 5 shows another example of data that a scope can provide. This
histogram of an eye-diagram's crossover point reveals a
distribution with two peaks. The twin peaks indicate deterministic
jitter, which comes from an outside source of interference such as
a switching power supply. You can often trace deterministic jitter
back to its source. "Jitter Fundamentals Brochure," listed in
"Jitter lessons," above, provides a good explanation of
deterministic jitter. The other type of jitter—random
jitter—follows a Gaussian distribution, and you can't determine its
source.
Figure 5 The twin peaks in this jitter histogram indicate deterministic
jitter. The distance between teh peaks indicates the deterministic
jitter's amplitude. Courtesy of Tektronix.
Do you need all three?When you need to equip a lab or production facility with
jitter-measuring equipment, you must decide how many of these
instruments to purchase. You may think you can forgo BER testers
and jitter analyzers and purchase just a scope. Because they take a
series of samples on a waveform, oscilloscopes can display more
information about a signal than BER testers and jitter analyzers
without scope displays can. By looking at the entire waveform, and
not just the edges, a scope can provide information about a
signal's amplitude as well as its timing.
On the downside, taking all those samples means that, when
measuring jitter, you'll get samples containing amplitude
information (although today's scopes contain sufficient memory for
many jitter-measurement applications). You won't get all that extra
information with jitter analyzers or BER testers, but these other
instruments give you more edge samples.
You'll hear conflicting claims from equipment manufacturers about
which instruments to use. BER tester manufacturers will tell you
that only their equipment gives you enough samples. Jitter-analyzer
manufacturers will tell you that scopes can't give you enough
samples to accurately measure jitter. Scope manufacturers will
argue that scopes have enough memory to do the job and that only
scopes let you see a signal's amplitude. Tim Margeson, product
manager at Tektronix (Beaverton, OR), states that "in some labs,
all three types of equipment have roles to play due to their unique
capabilities."
Recently, some test-equipment makers have developed hybrid
instruments. BER testers, which traditionally reported bit errors
only, now perform some jitter analysis, and some even include
sampling oscilloscopes. Jitter analyzers may now also contain
sampling oscilloscopes; one example is the Wavecrest SIA-3000 (see
Product Update ). These sampling scopes let you view eye diagrams
but they don't have the bandwidth that you get from stand-alone
sampling scopes. The scope bandwidth of a hybrid instrument
currently tops at 6 GHz. Real-time and equivalent-time sampling
scopes now offer software that measures jitter and estimates BER.
They key word here is "estimate," for you can get a true measure of
BER with a BER tester only.Manufacturers mentioned in this article
Agilent Technologies
Palo Alto, CA
800-452-4844
www.tm.agilent.com Amherst Systems
Amherst, MA
413-596-5354
www.amherst-systems.com
Anritsu
408-778-2000
Morgan Hill, CA
www.us.anritsu.com
GuideTech
Sunnyvale, CA
408-733-6555
www.jitter.com
LeCroy
Chestnut Ridge, NY
845-578-6020
www.lecroy.com
Synthesys Research
Menlo Park, CA
408-364-1853
www.synthesysresearch.com
Tektronix
Beaverton, OR
800-426-2200
www.tektronix.com
Wavecrest
Eden Prairie, MN
952-831-0030
www.wavecrest.com
Author InformationMartin Rowe has a BSEE from Worcester Polytechnic Institute and an
MBA from Bentley College. Before joining T&MW in 1992, he
worked for 12 years as a design engineer for manufacturers of
semiconductor process equipment and as an applications engineer for
manufacturers of measurement and control equipment. E-mail:
m.rowe@tmworld.com.
ReferencesRowe, Martin, "BER measurements reveal network health," Test &
Measurement World, July 2002. p.45. www.tmworld.com/archives .
Nelson, Rick, "Serdes devices challenge ATE," Test &
Measurement World, August 2002. p.19. www.tmworld.com/archives.
Not just a communications problemClock jitter affects more than just communications circuits and
systems; it can affect analog-to-digital converters (ADCs) and
other clocked circuits, too. ADCs rely on a clock to start a
conversion. If that clock's edges vary in time relative to each
other, then the ADC will sample at the wrong time. Depending on the
slope of the input signal, the jitter could cause an incorrect
digital output and the ADC will produce the wrong code.
Figure A
Figure BFigure A shows a hypothetical 3-bit ADC that converts a linearly
increasing voltage. In this example, the clock has no jitter and
the output codes are unambiguous, increasing linearly with the
input voltage. Figure B shows the effects of jitter on the clock
signal. The fourth sample, for instance, occurs at the transition
point from 010 to 011, and the ADC may return either value. In
addition, no sample point falls within the "100" bin, so the ADC
won't return that value.
Incorrect output codes will produce distortion in the digitized
signal that wasn't present in the original analog signal. You can
see the distortion if you perform a fast Fourier transform on the
digitized signal. By investigating the ADC's output codes, you can
infer the amount of jitter in the clock signal.
Jitter lessons on the Web"Analyzing Jitter Using a Spectrum Approach," Tektronix, Beaverton,
OR, 2002.
http://www.tektronix.com/Measurement/App_Notes/55_15631/eng/55W_15631_0.pdf
"Histograms Simplify Analysis of Random Jitter," Application note
AN1200-9, Agilent Technologies, Palo Alto, CA.
cp.literature.agilent.com/litweb/pdf/5966-4482E.pdf.
"Jitter Analysis Techniques Using an Agilent Infiniium
Oscilloscope," Product Note, Agilent Technologies, Palo Alto, CA,
2002. cp.literature.agilent.com/litweb/pdf/5988-6109EN.pdf.
"Jitter Fundamentals Brochure," Wavecrest, Eden Prairie, MN.
www.wavecrest.com/technical/jitterfund.htm.
"MP1580A Portable 2.5G/10G Analyzer," Application Note, Anritsu,
Morgan Hill, CA, 2001.
www.anritsu.co.jp/Products/pdf_e/MP1580A_EF1100.pdf .
Soo, Nelson, "Jitter Measurement Techniques," Application Brief
AB36, Pericom Semiconductor, San Jose, CA.
www.pericom.com/pdf/applications/AB36.pdf .
"Some timing/jitter/measurement-specific terms," Amherst Systems,
Amherst, MA. www.thejittersolution.com/glossary.htm.
"Understanding Jitter: Getting Started," Wavecrest, Eden Prairie,
MN, 2001.
www.wavecrest.com/technical/VISI_6_Getting_Started_Guides/6understanding.pdf.
BERT and scopeDavid Andres, an engineer at Marvell Semiconductor (Sunnyvale, CA),
finds that he needs both a BER tester and an oscilloscope to
measure jitter in his company's Serdes devices. These devices
transmit and receive 10 Gigabit Ethernet electrical signals.
Beginning with an Agilent 71612C BER tester, Andres runs a BER scan
to measure a transmitter's BER or a receiver's tolerance to jitter.
If a device fails to meet a 10–12 BER limit at speeds up to 13
Gbits/s, Andres pulls out a Tektronix TDS7404 oscilloscope to
investigate.
With the scope, Andres captures serial bit streams using about 8
Msamples of the scope's 32-Msample memory. From just 8 Msamples,
Andres gets enough data to identify the device's jitter components.
He uses analysis software developed in-house, along with software
from Amherst Systems (Amherst, MA) to measure jitter on a device's
clock signals. After downloading the captured waveform from the
scope to a PC, Andres plots the time difference from each clock
edge to an ideal clock. He then performs spectral analysis on the
result to look for clues as to the jitter's source.
;){kind=link}