SY87700/SY87701 CDR
EVALUATION KIT
SY87700/SY87701
EVALUATION BOARD
FEATURES
s
3.3V power supply:
Split V
CC
= +2V, GND = 0V
V
EE
= –1.3V for 3.3V
V
EE
= –3V for 5.0V
s
Simple switch configuration
s
PECL signal outputs
s
Simple RDIN+, RDIN– PECL inputs
s
Simple REFCLK TTL input
s
SY87700: Clock and data recovery from 32Mbps up
to 175Mbps NRZ data stream, clock generation from
32Mbps to 175Mbps
s
SY87701: Clock and data recovery from 32Mbps up
to 1.25Gbps NRZ data stream, clock generation from
32Mbps to 1.25Gbps
DESCRIPTION
The SY87700 and SY87701 Clock and Data Recovery
(CDR) chips are both high-performance ICs that are
designed to provide protocol-independent clock and data
recovery at any data rate between 32Mbps and 175Mbps
for the SY87700 and 32Mbps to 1.25Gbps for the SY87701.
This document provides design and implementation
information, as well as a detailed description of the SY87700/
701 evaluation board.
The evaluation board is intended to provide a convenient
test and evaluation platform for the SY87700/701 CDR
device. This board can be used for many types of jitter
tests, including SONET compliance of the SY87700/701,
as well as PLL characterization.
FUNCTIONAL BLOCK DIAGRAM
CH1
CH2
TRIG
CLKOUT
Scope
150 ps TTC*
150 ps TTC*
BERT Stack
DATAIN
DATAOUT–
CLKIN
V
CC
+2V
LED ON - LOCK
50Ω Term.
J1
J4
J5
50Ω Term.
J2
LFIN
PECL
RDIN+ (PECL)
RDIN— (PECL)
PECL
RDOUT+
RDOUT—
RCLK+
PECL
J10
PECL
J9
PECL
J12
PECL
J11
SY87700
SY87701
RCLK—
J6
250 ps TTC*
J3
50Ω Term.
TTL
TCLK+
REFCLK (TTL)
TCLK—
PECL
PECL
J8
J7
GND 0V
V
EE
(—1.3V for 3.3V)
(—3V for 5.0V)
Z
O
= 50Ω
32 EP-TQFP
Pulse
Generator
*Note: TTC = HP / Agilent Transition Time Converter
150ps:HP15435A
2000ps: HP15438A
Spectrum
Analyzer
Figure 1. SY87700/SY87701 Evaluation Board and Test Set-Up
Rev.: A
Amendment: /0
1
Issue Date: July 2002
Micrel
SY87700/701
Evaluation Board
FUNCTIONAL DESCRIPTION
The evaluation board simplifies test and measurement of
the SY87700 and SY87701. This section covers the various
parts of the SY87700/701 evaluation board, and includes
detailed information about these blocks. Performance of
the SY87700/701 can be easily evaluated by following the
step-by-step instructions found in the “Test Configuration”
section.
Power Supply
The SY87700L and SY87701L are 3.3V devices.
Therefore, V
CC
should all be connected to 2.0V, and GND
connected to 0V, and V
EE
should be connected to –1.3V.
The SY87700V and SY87701V are 5.0V devices, therefore,
V
CC
should be connected to 2V, and GND to 0V, and V
EE
should be connected to –3V.
Board Design and Layout
The evaluation board uses a force-sense design on the
signal inputs where the signal pins (source pins) on the
SY87700/701 are located on 50Ω line, on the last layer.
The sense lines, however, are located on layer 1. The force-
sense design is handy for monitoring inputs to the SY87700/
701 (such as input jitter). However, a 50Ω terminator needs
to be added to all unused sense outputs or the line will act
as a quarter wave stub notch filter.
LED
The SY87700/701 evaluation board features one LED
for monitoring the Link Fault Indicator (LFIN) pin. The LED
will turn on when the PLL has locked-on to the RDIN input
data stream, which indicates that LFIN has gone active
HIGH. Additionally, LFIN can only go active when CD is
HIGH and RDIN is within the 1000ppm frequency range of
the PLL.
Signal Inputs
Signal RDIN is 3.3V/5V PECL DC-coupled. Therefore,
the current level for DC-coupled applications is V
CC
–2V.
RDIN-DRIVEN
V
CC
V
CC
V
CC
+2V
J4
J5
RDIN-BERT
If you are using a high frequency bit error rate tester
(such as the Agilent 70843B Error Performance Analyzer)
to drive RDIN±, you will need to insert a 250ps Transition
Time Converter (TTC) to slow its edge down.
REFCLK
If you are using a high frequency clock or pulse generator
such as the Agilent 8133 to drive REFCLK, you will need to
insert a 2000ps Transition Time Converter (TTC) to slow its
edge down.
Signal Outputs
The SY87700/701 features PECL outputs for both
RDOUT± and RCLK± and TCLK±. Unused pins should be
left FLOATING.
Test Configuration
This section contains step-by-step instructions for
configuring the SY87700 and SY87701 for clock and data
from the data stream of a BERT stack.
1. Set switches on evaluation board for desired data and
clock frequencies. There are seven switches in SW1:
1. FREQSEL1
2. FREQSEL2
3. FREQSEL3
4. CLKSEL
5. DIVSEL2
6. DIVSEL1
7. CD
See “All Possible Legal Frequency and Divide Selec-
tions” section on page 5, on how to set these switches. In
addition, CLKSEL should be set HIGH which configures
TCLK output as the recovered CLK from RDIN. If CLKSEL
is low, TCLK will be the synthesized clock output. Addition-
ally, CD should be set HIGH to allow the PLL to recover
RDIN. If CD is low, RDIN is forced low.
2. Connect GND to 0V.
3. Connect VCC to +2V.
4. For 3.3V operation, connect VEE = –1.3V.
For 5.0V operation, connect VEE = –3.0V.
5. Connect REFCLK (TTL) inputs to reference clock.
Note: If using Agilent 8133A Pulse Generator, use
250ps Time Transistion Converters on the 8133
outputs.
6. Connect TCLK (PECL) outputs to data inputs on test
equipment.
7. Connect RDINV (PECL) inputs to data source.
8. Connect RDOUT (PECL) to outputs on test equipment.
9. Connect RCLK outputs to clock inputs on test
equipment.
R1
Z=50Ω
R1
RDIN+
RDIN–
Z=50Ω
R2
R2
GND 0V
Note: For +5V systems
R1 = 82Ω, R2 = 130Ω
For +3V systems
R1 = 150Ω, R2 = 75Ω
V
T
= V
CC
–2
V
EE
(–1.3V for 3.3V)
(–3V for 5.0V)
Figure 2. Test Set-Up
2
Micrel
SY87700/701
Evaluation Board
FREQUENTLY ASKED QUESTIONS
What Do I Do with the Exposed Pad on the Bottom of
the Package?
The purpose of the exposed pad at the bottom of the
package is to conduct heat more efficiently out of the
package. Solder or use thermal conductive epoxy. Although
the pad is connected to V
EE
, will not be any degradation in
either output generated jitter or input jitter tolerance
performance.
I Just Got my Evaluation Board and I Cannot Get
Anything to Work.
First check the power supplies. This evaluation board
uses one power supply. You should see a current draw of
about 200mA when the part is running normally. After that,
check voltage swing levels of REFCLK. It is important to
focus on getting the synthesizer (CMU) to work first (REFCLK
to TCLK), before the data recovery side. TCLK synthesizer
sets up the coarse adjust for the VCO in the CDR (or CRU),
so if TCLK is not oscillating at the right frequency, the CDR
will not lock. Another tip: use a frequency counter like
HP53132A to measure frequency of TCLK– it is often more
foolproof than using the DSO. If using a DSO scope, like
the Agilent CSA803, or the 11801 from Tektronix, trigger off
of the REFCLK clock source.
After the synthesizer is operating as expected, make sure
to change the trigger on the oscilloscope to trigger on the
data generation instrument, such as second HP8133A, a
Microwave Logic 1400, or HP70004A,70841 BERT stack.
The BERT stack has a “clock output”, that be used to trigger
the scope. The instrument generating REFCLK is not phase/
frequency locked to the data generation side, so it would be
impossible to examine an “eye” diagram.
Check the eye of the output source directly first, before
going into the device. Most data generation instruments
have deskew capability. It is important to deskew both the
instrument and the
±
coaxial cables into the DSO, otherwise
you’ll have too much apparent deterministic jitter.
Aside from setting the DIVSEL, and FREQSEL incorrectly,
everything should operate as expected at this point.
What is the Time Domain Reflectometry Test?
TDR (Time Domain Reflectometry) is used to verify
impedance continuity along a signal path. Many
interconnects, such as SMA, if not launched correctly onto
the PCB will exhibit inductive-like resonance with an abrupt
capacitive discontinuity. This discontinuity will subtract signal
from the inputs and outputs and effectively close the resulting
data eye.
What Should I Use to Generate REFCLK in My Design?
This depends on data rate, jitter budget, and cost.
However, REFCLK input jitter will affect the overall jitter
performance of the system. A fundamental tone crystal-
based oscillator is ideal. Measure the jitter of the oscillator
with a Wavecrest DTS2077. A measurement above the
3ps noise floor of the instrument is too high. Remember
that the REFCLK input is multiplied by the DIVSEL selected
value, so the resulting jitter increases by 20log (divide ratio).
If you use a clock derived from an ASIC, verify the single
cycle and accumulated cycle jitter.
Crystal based oscillators typically have poor AC power
supply rejection ratio, and if you are providing board power
via 400kHz switching supplies you may have to provide
some level of filtering, not just bypassing, for the supplies.
Also verify that the oscillator output has no “pedestals” in
the response due to improper impedance matching and/or
inadequate drive capability of the oscillator.
Do not use CMOS-based PLLs. They almost always have
too much high frequency deterministic jitter for this
application. Also fanning out one oscillator to several
locations on your board is not a good idea. Crosstalk and
inadequate drive can adversely affect performance. We
recommend Raltron, Mutron, CTS, Plantronics, Frequency
Management, etc., as vendors of crystal-based fundamental
tone oscillators.
Can you Suggest a Bypass/Decoupling Scheme?
The SY87700/701 data sheet contains the evaluation
board schematic, and a bill of materials list is included in
this document. We have found this arrangement to be an
excellent starting point. In addition, most system designs
could be dramatically improved by spacing the power planes
between ground planes to lower the self-inductance of the
power distribution.
3
Micrel
SY87700/701
Evaluation Board
What Layout Tips Do You Have?
1.
Establish controlled impedance stripline, microstrip,
or co-planar construction techniques for high-speed
signal paths.
2.
All differential paths are critical timing paths, and
skew should be matched to within
±10psec.
3.
Signal trace impedance should not vary more than
±5%.
If in doubt, perform TDR analysis of signal
traces.
4.
Maintain compact filter networks as close to filter
pins as possible.
5.
Provide ground plane relief under the filter path to
reduce stray capacitance and be careful of crosstalk
coupling into the filter network.
Maintain low jitter on the REFCLK input by isolating
the XTAL oscillator from power supply noise by
adequately decoupling.
Keep the XTAL oscillator close to SY87700/701.
High speed operation may require use of
fundamental-tone crystal-based oscillator for
optimum performance. (Third overtone oscillators
typically have more jitter.)
Isolate the input, output, and REFCLK signal traces
from other clock and data signals on your board if
these other traces are within 3x the trace width.
Isolation can be achieved by putting ground traces
in between.
6.
7.
8.
9.
How Do You Suggest We Qualify and Evaluate
Performance?
Evaluation should start by measuring the jitter of the
REFCLK input. The Clock Multiplier Unit (CMU) is simply a
PLL. It multiplies the incoming REFCLK frequency, and jitter
will usually worsen. The HP8133A pulse generator is ideal,
and the user should include a Transition Time Converter on
the 8133s output to slow its edges down. Make sure the
rise/fall times are reasonable (not 28ps rise/fall found on
the 12Gbps HP BERT clocks!) and 150ps TTCs will ensure
this. Measure the TCLK output jitter using either the
±
side,
with the other side terminated. Suitable instruments for
measuring the TCLK jitter are the CSA803, 11801, or the
Wavecrest 2077. See Figure 1 for descriptions of set-up.
Characterization of the jitter must include accumulation of
many cycles or periods down to a specified low pass corner
frequency. Wavecrest makes this easy with their 6.1 version
software since the user can specify a low pass corner for
the collected jitter. The Wavecrest instrument cannot be set
up for single period measurements, but must look at the
difference between the rising edges of the REFCLK and
the TCLK using both channels and performing a histogram
of the propagation time between the input REFCLK (which
is the HP8133A trigger divided by one) and the output TCLK.
Evaluation of the CDR is similar, except that the RCLK
and RDOUT outputs are used instead. The procedure for
measuring the RCLK jitter is identical to the above procedure
for TCLK jitter.
Evaluation of the output jitter on RDOUT using RCLK as
a trigger source isn’t trivial, as the minimum time between
the scope trigger and measurement is 24ns for the Agilent
86100A scope. Therefore the user must delay the data by
the same amount, so that the jitter on RDOUT is measured
with respect to the correct clock edge. This is important, as
the SY87700/701 will retime the edges on RDOUT so that
they better align with RCLK. The Wavecrest DTS2077 can
also be used.
The setup for SONET jitter compliance tests is shown in
Figure 1. Agilent provides software for automated Bellcore
jitter compliance tests. Contact Agilent for details.
Should I Adjust the Loop Lilter?
The values found in the data sheets are the result of
extensive modeling as well as lab testing. Therefore, we
recommend starting with those values. Selecting values to
simply reduce jitter does not work since there is a trade-off
in jitter generation and jitter tolerance. However, for telecom
applications under Bellcore,ITU/CCIT specifications it may
be advantageous to adjust the values to trade off jitter
transfer for jitter generation.
4
Micrel
SY87700/701
Evaluation Board
DESCRIPTION OF CONNECTORS
Connects to
28-pin SOIC
Pin 4
Pin 5
Pin 7
Pin 4
Pin 5
Pin 7
Pin 18
Pin 19
Pin 21
Pin 22
Pin 24
Pin 25
Connector
J1
J2
J3
J4
J5
J6
J7
J8
J9
J10
J11
J12
Name
RDIN+_S
RDIN–_S
REFCLK–S
RDIN+_F
RDIN–_F
REFCLK_F
TCLK–
TCLK+
RCLK–
RCLK+
RDOUT–
RDOUT+
Type
PECL
PECL
TTL
PECL
PECL
PECL
PECL
PECL
PECL
PECL
PECL
PECL
32-pin TQFP
Pin 2
Pin 3
Pin 5
Pin 2
Pin 3
Pin 5
Pin 17
Pin 18
Pin 20
Pin 21
Pin 23
Pin 24
Description
RDIN+ (Sense)
RDIN– (Sense)
REFCLK– (Sense)
RDIN+ (Force)
RDIN– (Force)
REFCLK– (Force)
TCLK– (Output)
TCLK+ (Output)
RCLK– (Output)
RCLK+ (Output)
RDOUT– (Output)
RDOUT+ (Output)
ALL POSSIBLE LEGAL FREQUENCY AND DIVIDER SELECTIONS
FREQSEL1
0
1
1
1
1
0
0
FREQSEL2
1
0
0
1
1
1
0
FREQSEL3
1
0
1
0
1
0
X
(2)
f
VCO
/f
RCLK
6
8
12
16
24
—
—
f
RCLK
Data Rates (Mbps)
125 –175
94 – 157
63 – 104
47 – 78
32 – 52
undefined
undefined
NOTES:
1. SY87700L operates from 32-175MHz. For higher speed applications, the SY87701L operates from 32-1250MHz.
2. X is a DON'T CARE.
DIVSEL1
0
0
1
1
DIVSEL2
0
1
0
1
f
RCLK
/ f
REFCLK
8
10
16
20
Table 1. M-Divider, f
RCLK
/ f
REFCLK
Divider Setting
5