1. Any-Frequency Precision Clock Product Family Overview
Silicon Laboratories Any-Frequency Precision Clock products provide jitter attenuation and clock multiplication/
clock division for applications requiring sub 1 ps rms jitter performance. The device product family is based on
Silicon Laboratories' 3rd generation DSPLL technology, which provides any-frequency synthesis and jitter
attenuation in a highly integr ated PLL solution that eliminates the need for discrete VCXO/VCSOs and loop filter
components. These devices are ideally suited for applications which require low jitter reference clocks, including
OTN (OTU-1, OTU-2, OTU-3, OTU-4), OC-48/STM-16, OC-192/STM-64, OC-768/STM-256, GbE, 10GbE, Fibre
Channel, 10GFC, synchronous Ethernet, wireless backhaul, wireless point-point infrastructure, broadcast video/
HDTV (HD SDI, 3G SDI), test and measurement, data acquisition systems, and FPGA/ASIC reference clocking.
Table 1 provides a product selector guide for the Silicon Laboratories Any-Frequency Precision Clocks. Three
product families are available. The Si5316, Si5319, Si5323, Si5324, Si5326, Si5366, and Si5368 are jitterattenuating clock multipliers that provide ultra-low jitter generation as low as 0.30 ps RMS. The devices vary
according to the number of clock inputs, number of clock outputs, and control method. The Si5316 is a fixedfrequency, pin controlled jitter attenuator that can be used in clock smoothing applications. The Si5323 and Si5366
are pin-controlled jitter-attenuating clock multipliers. The frequency plan for these pin-controlled devices is
selectable from frequency lookup tables and includes common frequency translations for SONET/SDH, ITU G.709
Forward Error Correction (FEC) applications (255/238, 255/237, 255/236, 238/255, 237/255, 236/255), Gigabit
Ethernet, 10G Ethernet, 1G/2G/4G/8G/10G Fibre Channel, ATM and broadcast video (Genlock). The Si5319,
Si5324, Si5326, Si5327, Si5368, and Si5369 are microprocessor-controlled devices that can be controlled via an
2
C or SPI interface. These microprocessor-controlled devices accept clock inputs ranging from 2 kHz to 710 MHz
I
and generate multiple independent, synchronous clock outputs ranging from 2 kHz to 945 MHz and select
frequencies to 1.4 GHz. Virtually any frequency translation combination across this operating range is supported.
Independent dividers are available for every input clock and output clock, so the Si5324, Si5326, Si5327, and
Si5368 can accept input clocks at different frequencies and generate output clocks at different frequencies. The
Si5316, Si5319, Si5323, Si5326, Si5327, Si5366, Si5368, and Si5369 support a digitally programmable loop
bandwidth that can range from 60 Hz to 8.4 kHz. An exter nal (37–41 MHz, 55–61 MHz, 109–125.5 MHz, or 163–
180 MHz) reference clock or a low-cost 114.285 MHz 3rd overtone crystal is required for these devices to enable
ultra-low jitter generation and jitter attenuation. (See "Appendix A—Narrowband References" on page 119.) The
Si5324 and Si5369 are much lower bandwidth devices, providing a user-programmable loop bandwidth from 4 to
525 Hz.
The Si5323, Si5324, Si5326, Si5327, Si5366, Si5368, and Si5369 support hitless switching between input clocks in
compliance with GR-253-CORE and GR-1244-CORE that greatly minimizes the prop ag ation of phase transie nt s to
the clock outputs during an input clock transition (<200 ps typ). Manual, automatic revertive and automatic nonrevertive input clock switching options are available. The devices monitor the input clocks for loss-of-signal and
provide a LOS alarm when missing pulses on any of the input clocks are detected. The devices monitor the lock
status of the PLL and provide a LOL alarm when the PLL is unlocked. The lock detect algorithm works by
continuously monitoring the phase of the selected input clock in relation to the phase of the feedback clock. The
Si5326, Si5366, Si5368, and Si5369 monitor the frequency of the input clocks with respect to a reference
frequency applied to an input clock or the XA/XB input, and generates a frequency offset alarm (FOS) if the
threshold is exceeded. This FOS feature is available for SONET/SDH applications. Both Stratum 3/3E and SONET
Minimum Clock (SMC) FOS thresholds are supported.
The Si5319, Si5323, Si5324, Si5326, Si5366, Si5368, and Si5369 provide a digital hold capability that allows the
device to continue generation of a stable output clock when the selected input reference is lost. During digita l hold,
the DSPLL generates an output frequency based on a historical average that existed a fixed amount of time before
the error event occurred, eliminating the effects of phase and frequency transients that may occur immediately
preceding entry into digital hold.
The Si5322, Si5325, Si5365, and Si5367 are frequency flexible, low jitter clock multipliers that provide jitter
generation of 0.6 ps RMS without jitter attenuation. These devices provide low jitter integer clock multiplication or
fractional clock synthesis, but they are not as frequency-flexible as the Si5319/23/24/26/66/68/69. The devices
vary according to the number of clock inputs, number of clock ou tput s, and co ntrol method . The Si5322 and Si5365
are pin-controlled clock multipliers. The frequency plan for these devices is selectable from frequency lookup
tables.
12Rev. 0.5
Si53xx-RM
A wide range of settings are available, but they are a subset of the frequency plans supported by the Si5323 and
Si5366 jitter-attenuating clock multipliers. The Si5325 and Si5367 are microprocessor-controlled clock multipliers
that can be controlled via an I
These devices accept clock inputs ranging from 10 MHz to 710 MHz and generate multiple independent,
synchronous clock outputs ranging from 10 MHz to 945 MHz and select frequencies to 1.4 GHz. The Si5325 and
Si5367 support a subset of the frequen cy translations available in the Si5319, Si5 324, Si5326, Si5327, Si5 368, and
Si5369 jitter-attenuating clock multipliers. The Si5325 and Si5367 can accept input clocks at different frequencies
and generate output clocks at different frequencies. The Si5322, Si5325, Si5365, and Si5367 support a digitally
programmable loop bandwidth that ranges from 150 kHz to 1.3 MHz. No external components are required for
these devices. LOS and FOS monitoring is available for these devices, as described above.
The Si5374 and Si5375 are quad DSPLL versions of the Si5324 and Si5319, respectively. Each of the four
DSPLLs can operate at completely independent frequencies. The only resources that they share are a common
2
C bus and a common XA/XB jitter reference oscillator. The Si5375 consists of four one-input and one-output
I
DSPLLs. The Si5374 consists of four two-input and two-output DSPLLs with very low loop bandwidth.
The Any-Frequency Precision Clocks have dif fere ntial clock ou tp ut(s) with pr ogr ammab le sign al for mats to support
LVPECL, LVDS, CML, and CMOS loads. If the CMOS signal format is selected, each differential output buffer
generates two in-phase CMOS clocks at the same frequency. For system-level debugging, a PLL bypass mode
drives the clock output directly from the selected input clock, bypassing the internal PLL.
Silicon Laboratories offers a PC-based software utility, DSPLLsim, tha t can be used to determine valid fre quency
plans and loop bandwidth settings for the Any-Frequency Precision Clock product family. For the microprocessorcontrolled devices, DSPLLsim provides the optimum PLL divider settings for a given input frequency/clock
multiplication ratio combination that minimizes phase noise and power consumption. Two DSPLLsim configuration
software applications are available for the 1-PLL and 4-PLL devices, respectively. DSPLLsim can also be used to
simplify device selection and configuration. This utility can be downloaded from http://www.silabs.com/timing.
Other useful documentation, including device data sheets and programming files for the microprocessor-controlled
devices are available from this website.
Si53222270710500.6 ps rms typ
Si532522
Si53654570710500.6 ps rms typ
Si536745
Notes:
1. Maximum input and output rates may be limited by speed rating of device. See each device’s data sheet for ordering
information.
2. Requires external low-cost, fixed frequency 3rd overtone 114.285 MHz crystal or reference clock. See "T able 60.X A/XB
Reference Sources and Frequencies" on page 119.
Clock Inputs
Clock Outputs
P Control
Max Input Freq (MHz)
Max Output Frequency (MHz)
Jitter Generation
(12 kHz – 20 MHz)
LOS
Hitless Switching
FOS Alarm
LOL Alarm
FSYNC Realignment
36 Lead 6 mm x 6 mm QFN
100 Lead 14 x 14 mm TQFP
71014000.6 ps rms typ
71014000.6 ps rms typ
1.8, 2.5, 3.3 V Operation
1.8, 2.5 V Operation
Rev. 0.515
Si53xx-RM
2. Narrowband vs. Wideband Overview
The narrowband (NB) devices offer a number of features and capabilities that are not available with the wideband
(WB) devices, as outlined in the below list:
Broader set of frequency plans due to more divisor options
Hitless switching between input clocks
Lower minimum input clock frequency
Lower loop bandwidth
Digital Hold (reference-based holdover instead of VCO freeze)
FRAMESYNC realignment
CLAT and FLAT (input to output skew adjust)
INC and DEC pins
PLL Loss of Lock status indicator
FOS is not supported.
16Rev. 0.5
Si53xx-RM
2
DSPLL
®
C1B
CS
LOL
BWSEL[1:0]
DBL_BY
Xtal or Refclock
SFOUT[1:0]
CKOUT+
CKOUT–
CKIN_1+
CKIN_1–
CKIN_2+
CKIN_2–
Control
Signal
Detect
VDD
GND
Frequency
Control
Bandwidth
Control
C2B
2
FRQSEL[1:0]
RST
0
1
RATE[1:0]
XA
XB
f
OSC
2
0
1
÷ N31
÷ N32
f
3_1
f
3_2
CK1DIV
CK2DIV
f
3
3. Any-Frequency Clock Family Members
3.1. Si5316
The Si5316 is a low jitter, precision jitter attenuator for high-speed communication systems, including OC -48, OC192, 10G Ethernet, and 10G Fibre Channel. The Si5316 accepts dual clock inputs in the 19, 38, 77, 155, 311, or
622 MHz frequency range and generates a jitter-attenuated clock output at the same frequency. Within each of
these clock ranges, the device can be tuned approximately 14% higher tha n nomin al SONET/SDH fre quencies, up
to a maximum of 710 MHz in the 622 MHz range. Th e DSPL L lo op b an dw idth is digitally selec table, providing jitter
performance optimization at the application level. Operating fr om a single 1.8, 2.5, or 3.3 V supp ly, the Si5316 is
ideal for providing jitter attenuation in high performance timing applications. See "6. Pin Control Parts (Si5316,
Si5322, Si5323, Si5365, Si5366)" on page 50 for a complete description.
The Si5319 is a jitter-attenuating precision M/N clock multiplier for applications requiring sub 1 ps jitter
performance. The Si5319 accepts one clock input ranging from 2 kHz to 710 MHz and generates one clock output
ranging from 2 kHz to 945 MHz and select frequencies to 1.4 GHz. The Si5319 can also use its crystal oscillator as
a clock source for frequency synthesis. The device provides virtua lly any frequency translation combination across
this operating range. The Si5319 input clock frequency and clock multiplication r atio are programmable through an
I2C or SPI interface. The Si5319 is based on Silicon Laboratories' 3rd-generation DSPLL
provides any-frequency synthesis and jitter attenuation in a highly integrated PLL solution that eliminates the need
for external VCXO and loop filter components. The DSPLL loop bandwidth is digitally programmable, providing
jitter performance optimization at the application level. Operating from a single 1.8, 2.5, or 3.3 V supply, the Si5319
is ideal for providing clock multiplication and jitter attenuation in high performance timing applications. See "7.
Microprocessor Controlled Parts (Si5319, Si5324, Si5325, Si5326, Si5327, Si5367, Si5368, Si5369, Si5374,
Si5375)" on page 76 for a complete description.
The Si5322 is a low jitter, precision clock multiplier for applications requiring clock multiplication without jitter
attenuation. The Si5322 acce pts dual clock inputs ranging from 19.44 to 707 MHz and generates two frequencymultiplied clock outputs ranging from 19.44 to 1050 MHz. The input clock frequency and clock multiplication ratio
are selectable from a table of popular SONET, Ethernet, Fibre Channel, and broadcast video (HD SDI, 3G SDI)
rates. The DSPLL loop bandwidth is digitally selectable from 150 kHz to 1.3 MHz. Operating from a single 1.8, 2.5,
or 3.3 V supply, the Si5322 is ideal for providing low jitter clock multiplication in high performance timing
applications. See "6. Pin Control Parts (Si5316, Si5322, Si5323, Si5365, Si5366)" on page 50 for a complete
description.
The Si5323 is a jitter-attenuating precision clock multiplier for high-speed communication systems, including
SONET OC-48/OC-192, Ethernet, Fibre Channel, and broadcast video (HD SDI, 3G SDI). The Si5323 accepts
dual clock inputs ranging from 8 kHz to 707 MHz and generates two frequency-multiplied clock outputs ranging
from 8 kHz to 1050 MHz. The input clock frequency and clock multiplication ratio are selectable from a table of
popular SONET, Ethernet, Fibre Channel, and broadcast video rates. The DSPLL loop bandwidth is digitally
selectable, providing jitter performance optimization at the application level. Operating from a single 1.8, 2.5, or
3.3 V supply, the Si5323 is ideal for providing clock multiplication and jitter attenuation in high-performance timing
applications. See "6. Pin Control Parts (Si5316, Si5322, Si5323, Si5365, Si5366)" on page 50 for a complete
description.
The Si5324 is a jitter-attenuating precision clock multiplier for applications requiring sub 1 ps jitter performance.
The Si5324 accepts dual clock inputs ranging from 2 kHz to 710 MHz and generates two independent,
synchronous clock outputs ranging from 2 kHz to 945 MHz and select frequencies to 1.4 GHz. The device provides
virtually any frequency translation combination across this operating range. The Si5324 input clock frequency and
clock multiplication ratios are programmable th rough an I
programmable, providing jitter performance optimization at the application level. The Si5324 features loop
bandwidth values as low as 4 Hz. Operating from a single 1.8, 2.5, or 3.3 V supply, the Si5324 is ideal for providing
clock multiplication and jitter attenuation in high-performance timing applications. See "7. Microprocessor
Controlled Parts (Si5319, Si5324, Si5325, Si5326, Si5327, Si5367, Si5368, Si5369, Si5374, Si5375)" on page 76
for a complete description.
2
C or SPI interface. The DSPLL loop bandwid th is digit ally
Figure 5. Si5324 Clock Multiplier and Jitter Attenuator Block Diagram
Rev. 0.521
Si53xx-RM
÷ N31
INT_C1B
÷ NC1
÷ NC2
Signal
Detect
C2B
0
1
CKOUT_2 +
CKOUT_2 –
CKOUT_1 +
CKOUT_1 –
/
/
2
2
1
0
1
0
SDA_SDO
RST
SCL
Control
SDI
A[2]/SS
A[1:0]
CMODE
CKIN_1 +
CKIN_1 –
2
2
CKIN_2 +
CKIN_2 –
÷ N32
0
1
BYPASS
÷ N2
f
3
DSPLL
®
VDD
GND
÷ N1_HS
f
OSC
3.6. Si5325
The Si5325 is a low jitter, precision clock multiplier for applications requiring clock multiplication without jitter
attenuation. The Si5325 accepts dual clock inputs ranging from 10 to 710 MHz and generates two independent,
synchronous clock outputs ranging from 2 kHz to 945 MHz and select frequencies to 1.4 GHz. The Si5325 input
clock frequency and clock multiplication ratios are programmable through an I
bandwidth is digitally program mable from 150 kHz to 1.3 MHz. Operating from a single 1.8, 2.5, or 3.3 V supply, the
Si5325 is ideal for providing clock multiplication in high performance timing applications. See "7. Microprocessor
Controlled Parts (Si5319, Si5324, Si5325, Si5326, Si5327, Si5367, Si5368, Si5369, Si5374, Si5375)" on page 76
for a complete description.
The Si5326 is a jitter-attenuating precision clock multiplier for applications requiring sub 1 ps jitter performance.
The Si5326 accepts dual clock inputs ranging from 2 kHz to 710 MHz and generates two independent,
synchronous clock outputs ranging from 2 kHz to 945 MHz and select frequencies to 1.4 GHz. The device provides
virtually any frequency translation combination across this operating range. The Si5326 input clock frequency and
clock multiplication ratios are programmable th rough an I
programmable from 60 Hz to 8 kHz, providing jitter performance optimization at the application level. Operating
from a single 1.8, 2.5, or 3.3 V suppl y, the Si5326 is ideal for providing clock multiplication and jitter atte nuation in
high-performance timing applications. See "7. Microprocessor Controlled Parts (Si5319, Si5324, Si5325, Si5326,
Si5327, Si5367, Si5368, Si5369, Si5374, Si5375)" on page 76 for a complete description.
2
C or SPI interface. The DSPLL loop bandwid th is digit ally
Figure 7. Si5326 Clock Multiplier and Jitter Attenuator Block Diagram
Rev. 0.523
Si53xx-RM
÷ N31
INT_C1B
Xtal or Re fc lo c k
÷ NC1
÷ NC2
Signal
Detect
VDD
GND
C2B
0
1
f
3
CKOUT_2 +
CKOUT_2 –
CKOUT_1 +
CKOUT_1 –
/
/
2
2
1
0
1
0
f
OSC
RATE[1:0 ]
LOL
CS_CA
SDA_SDO
INC
DEC
RST
SCL
Control
SDI
A[2]/SS
A[1:0]
XAXB
CMODE
CKIN_ 1 +
CKIN_ 1 –
2
2
CKIN_ 2 +
CKIN_ 2 –
÷ N32
0
1
3
BYPASS
÷ N2
DSPLL
÷ N1_HS
DSPLL
®
3.8. Si5327
The Si5327 is a jitter-attenuating precision clock multiplier for applications requiring sub 1 ps jitter performance.
The Si5327 accepts dual clock inputs ranging from 2 kHz to 710 MHz and generates two independent,
synchronous clock outputs ranging from 2 kHz to 808 MHz. The device provides virtually any frequency translation
combination across this operating range. The Si5327 input clock frequency and clock multiplication ratios are
programmable through an I
jitter performance optimization at the application level. The Si5327 features loop bandwidth values as low as 4 Hz.
Operating from a single 1.8, 2. 5, or 3.3 V supply, the Si5327 is ideal for providing clock multiplication and jitter
attenuation in high-performance timing applications. See "7. Microprocessor Controlled Parts (Si5319, Si5324,
Si5325, Si5326, Si5327, Si5367, Si5368, Si5369, Si5374, Si5375)" on page 76 for a complete description.
2
C or SPI interface. The DSPLL loop bandwidth is digitally programmable, providing
Figure 8. Si5327 Clock Multiplier and Jitter Attenuator Block Diagram
24Rev. 0.5
Si53xx-RM
C2A
CS0_C3A
C2B
CS1_C4A
ALRMOUT
C1A
CKIN_1+
CKIN_1–
CKIN_2+
CKIN_2–
C3B
CKIN_3+
CKIN_3–
CKIN_4+
CKIN_4–
C1B
VDD
GND
CKOUT_1+
CKOUT_1–
÷ NC1
1
0
CKOUT_2+
CKOUT_2–
÷ NC2
1
0
CKOUT_3+
CKOUT_3–
÷ NC3
1
0
CKOUT_4+
CKOUT_4–
÷ NC4
1
0
2
2
2
2
2
2
2
2
f
OSC
f
3
DBL2_BY
DBL34
DBL5
BWSEL[1:0]
FRQSEL[3:0]
DIV34[1:0]
FOS_CTL
SFOUT[1:0]
RST
CMODE
AUTOSEL
BYPASS/
DSBL2
Control
÷ N3_2
÷ N3_1
÷ N3_3
÷ N3_4
CKOUT_5+
CKOUT_5–
÷ NC5
1
0
2
FRQTBL
DIV34[1:0]
÷ N1_HS
DSPLL
®
÷ N2
3.9. Si5365
The Si5365 is a low jitter, precision clock multiplier for applications requiring clock multiplication without jitter
attenuation. The Si5365 accepts four clock inputs ranging from 19.44 MHz to 707 MHz and generates five
frequency-multiplied clock outputs ranging from 19.44 MHz to 1050 MHz. The input clock frequency and clock
multiplication ratio are selectable from a table of popular SONET, Ethernet, Fibre Channel, and broadcast video
rates. The DSPLL loop bandwidth is digitally selectable. Operating from a single 1.8, 2.5 V, or 3.3 V supply, the
Si5365 is ideal for providing clock multiplication in high performance timing applications. See "6. Pin Control Parts
(Si5316, Si5322, Si5323, Si5365, Si5366)" on page 50 for a complete descr iption.
The Si5366 is a jitter-attenuating precision clock multiplier for high-speed communication systems, including
SONET OC-48/OC-192, Etherne t, and Fibr e Chan nel. T he S i53 66 acce pts four clo ck inp uts ranging fr om 8 kHz to
707 MHz and generates five frequency-multiplied clock outputs ranging from 8 kHz to 1050 MHz. The input clock
frequency and clock multiplication ratio are selectable from a table of popular SONET, Ethernet, Fibre Channel,
and broadcast video (HD SDI, 3G SDI) rates. The DSPLL loop bandwidth is digitally selectable from 60 Hz to
8 kHz, providing jitter performance optimization at the application level. Operating from a single 1.8, 2.5, or 3.3 V
supply, the Si5366 is ideal for providing clock multiplication and jitter attenuation in high performance timing
applications. See "6. Pin Control Parts (Si5316, Si5322, Si5323, Si5365, Si5366)" on page 50 for a complete
description.
The Si5367 is a low jitter, precision clock multiplier for applications requiring clock multiplication without jitter
attenuation. The Si5367 accepts four clock inputs ranging from 10 to 707 MHz and generates five frequencymultiplied clock outputs ranging from 2 kHz to 945 MHz and select frequencies to 1.4 GHz. The Si5367 input clock
frequency and clock multiplication ratio are programmable through an I
bandwidth is digitally program mable from 150 kHz to 1.3 MHz. Operating from a single 1.8, 2.5, or 3.3 V supply, the
Si5367 is ideal for providing clock multiplication in high performance timing applications. See "7. Microprocessor
Controlled Parts (Si5319, Si5324, Si5325, Si5326, Si5327, Si5367, Si5368, Si5369, Si5374, Si5375)" on page 76
for a complete description.
2
C or SPI interface. The DSPLL loop
Figure 11. Si5367 Clock Multiplier Block Diagram
Rev. 0.527
Si53xx-RM
C2A
CS0_C3A
C2B
CS1_C4A
INT_ALM
C1A
CKIN_1+
CKIN_1–
CKIN_2+
CKIN_2–
C3B
CKIN_3+
CKIN_3–
CKIN_4+
CKIN_4–
C1B
CKIN_3
CKIN_4
CKOUT_2
VDD
GND
CKOUT_1+
CKOUT_1–
÷ NC1
1
0
CKOUT_2+
CKOUT_2–
÷ NC2
1
0
CKOUT_3+
CKOUT_3–
÷ NC3
1
0
CKOUT_4+
CKOUT_4–
÷ NC4
1
0
2
2
2
2
2
2
2
2
f
OSC
Xtal or Refc lock
RATE[1:0]
XA
XB
f
x
DSBL2/BYPASS
DSBL34
DSBL5
FSYNC
LOGIC/
ALIGN
SDA_SDO
SCL
SDI
A[1:0]
INC
DEC
FS_ALIGN
RST
CMODE
BYPASS/DSBL2
LOL
Control
÷ N3_2
÷ N3_1
FSYNC
÷ N2
÷ N3_3
÷ N3_4
CKOUT_5+
CKOUT_5–
÷ NC5
1
0
2
1
0
A[2]/SS
÷ N1_HS
DSPLL
®
f
3
3
3.12. Si5368
The Si5368 is a jitter-attenuating precision clock multiplier for applications requiring sub 1 ps rms jitter
performance. The Si5368 accepts four clock inputs ranging from 2 kHz to 710 MHz and generates five
independent, synchronous clock outputs ranging from 2 kHz to 945 MHz and select frequencies to 1.4 GHz. The
device provides virtually any frequency translation combination across this operating range. The Si5368 input clock
frequency and clock multiplication ratio are programmable through an I
bandwidth is digitally programmable from 60 Hz to 8 kHz, providing jitter performance optimization at the
application level. Operating from a single 1.8, 2.5, or 3.3 V supply, the Si5368 is ideal for providing clock
multiplication and jitter attenuation in high performance timing applications. See "7. Microprocessor Controlled
Parts (Si5319, Si5324, Si5325, Si5326, Si5327, Si5367, Si5368, Si5369, Si5374, Si5375)" on page 76 for a
complete description.
2
C or SPI interface. The DSPLL loop
Figure 12. Si5368 Clock Multiplier and Jitter Attenuator Block Diagram
28Rev. 0.5
Si53xx-RM
C2A
CS0_C3A
C2B
CS1_C4A
INT_ALM
C1A
CKIN_1+
CKIN_1–
CKIN_2+
CKIN_2–
C3B
CKIN_3+
CKIN_3–
CKIN_4+
CKIN_4–
C1B
CKIN_3
CKIN_4
CKOUT_2
VDD
GND
CKOUT_1+
CKOUT_1–
÷ NC1
1
0
CKOUT_2+
CKOUT_2–
÷ NC2
1
0
CKOUT_3+
CKOUT_3–
÷ NC3
1
0
CKOUT_4+
CKOUT_4–
÷ NC4
1
0
2
2
2
2
2
2
2
2
f
OSC
Xtal or Refclock
RATE[1:0]
XA
XB
f
x
DSBL2/BYPASS
DSBL34
DSBL5
FSYNC
LOGIC/
ALIGN
SDA_SDO
SCL
SDI
A[1:0]
INC
DEC
FS_ALIGN
RST
CMODE
BYPASS/DSBL2
LOL
Control
÷ N3_2
÷ N3_1
FSYNC
÷ N2
÷ N3_3
÷ N3_4
CKOUT_5+
CKOUT_5–
÷ NC5
1
0
2
1
0
A[2]/SS
÷ N1_HS
DSPLL
®
f
3
3
3.13. Si5369
The Si5369 is a jitter-attenuating precision clock multiplier for applications requiring sub 1 ps rms jitter
performance. The Si5369 accepts four clock inputs ranging from 2 kHz to 710 MHz and generates five
independent, synchronous clock outputs ranging from 2 kHz to 945 MHz and select frequencies to 1.4 GHz. The
device provides virtually any frequency translation combination across this operating range. The Si5369 input clock
frequency and clock multiplication ratio are programmable through an I
bandwidth is digitally programmable, providing loop bandwidth values as low as 4 Hz. Operating from a single 1.8,
2.5, or 3.3 V supply, the Si5369 is ideal for providing clock multiplication and jitter attenuation in high performance
timing applications. See "7. Microprocessor Controlled Parts (Si5319, Si5324, Si5325, Si5326, Si5327, Si5367,
Si5368, Si5369, Si5374, Si5375)" on page 76 for a complete description.
2
C or SPI interface. The DSPLL loop
3.14. Si5374/75 Compared to Si5324/19
In general, the Si5374 can be viewed as a quad version of the Si5324 and the Si5375 can be viewed as a quad
version of the Si5319. However, there are not exactly the same. This is an overview of the difference s:
1. The Si5374/75 cannot use a crystal as its OSC reference. It requires the use of a single external single-ended
or differential crystal oscillator.
Figure 13. Si5369 Clock Multiplier and Jitter Attenuator Block Diagram
2. The Si5374/75 only supports I
available on the Si5374/75.
3. The Si5374/75 does not provide separate INT_CK1B and CK2B pins to indicate wh en CKIN1 and CKIN2 do not
have valid clock inputs. Instead, the IRQ pin can be programme d to function as one pin, the other pin or both.
4. Selection of the OSC frequency is done by a register (RATE_REG), not by using the RATE pins.
5. The Si5374/75 uses a different version of DSPLLsim: Si537xDSPLLsim.
6. The Si5374/75 does not support 3.3 V operation.
2
C as its serial port protocol and does not have SPI. No I2C address pins are
Rev. 0.529
Si53xx-RM
CKIN3P_B
CKOUT3N_B
÷ N31
DSPLL
®
B
÷ NC1
÷ NC2
CKIN3N_B
CKIN4P_B
÷ N32
CKIN4N_B
Internal
Osc
PLL Bypass
CKOUT3P_B
CKOUT4N_B
CKOUT4P_B
f
OSC
÷ NC1_HS
Input
Monitor
Hitless
Switch
PLL Bypass
f
3
÷ N2
Status / Control
PLL Bypass
High PSRR
Voltage Regulator
VDD_q
GND
Synthesis Stage
CKIN1P_A
CKOUT1N_A
÷ N31
DSPLL
®
A
÷ NC1
÷ NC2
CKIN1N_A
CKIN2P_A
÷ N32
CKIN2N_A
Internal
Osc
PLL Bypass
CKOUT1P_A
CKOUT2N_A
CKOUT2P_A
Output Stage
f
OSC
÷ NC1_HS
Input
Monitor
Hitless
Switch
PLL Bypass
f
3
÷ N2
PLL Bypass
Input Stage
CKIN7P_D
CKOUT7N_D
÷ N31
DSPLL
®
D
÷ NC1
÷ NC2
CKIN7N_D
CKIN8P_D
÷ N32
CKIN8N_D
Internal
Osc
PLL Bypass
CKOUT7P_D
CKOUT8N_D
CKOUT8P_D
f
OSC
÷ NC1_HS
Input
Monitor
Hitless
Switch
PLL Bypass
f
3
÷ N2
PLL Bypass
CKIN5P_C
CKOUT5N_C
÷ N31
DSPLL
®
C
÷ NC1
÷ NC2
CKIN5N_C
CKIN6P_C
÷ N32
CKIN6N_C
Internal
Osc
PLL Bypass
CKOUT5P_C
CKOUT6N_C
CKOUT6P_C
f
OSC
÷ NC1_HS
Input
Monitor
Hitless
Switch
PLL Bypass
f
3
÷ N2
PLL Bypass
RSTL_q
CS_q
SCL SDA
LOL_q
IRQ_q
Low Jitter
XO or Clock
OSC_P/N
3.15. Si5374
The Si5374 is a highly integrated , 4- PLL jit ter- atten uat ing precision clock multiplier for applications requiring sub 1
ps jitter performance. Each of the DSPLL® clock multiplier engines accepts two input clocks ranging from 2 kHz to
710 MHz and generates two independent, synchronous output clocks ranging from 2 kHz to 808 MHz. Each
DSPLL provides virtually any frequency translation across this operating range. For asynchronous, free-running
clock generation applications, the Si5374’s reference oscillator can be used as a clock source for the four DSPLLs.
The Si5374 input clock frequency and clock multiplication ratio are programmable through an I2C interface. The
Si5374 is based on Silicon Laboratories' 3rd-generation DSPLL® technology, which provides any-frequency
synthesis and jitter attenuation in a highly integrated PLL solution that eliminates the need for external VCXO and
loop filter components. Each DSPLL loop bandwidth is digitally programmable from 4 to 525 Hz, providing jitter
performance optimization at the application level. The device operates from a single 1.8 or 2.5 V supply with onchip voltage regulators with excellent PSRR. The Si5374 is ideal for providing clock multiplication and jitter
attenuation in high port count optical line cards requiring independent timing domains.
Figure 14. Si5374 Functional Block Diagram
30Rev. 0.5
Si53xx-RM
CKIN1P_B
÷ N31
DSPLL
®
B
CKIN1N_B
÷ N32
f
OSC
÷ NC1_HS
Input
Monitor
f
3
÷ N2
Status / Control
PLL Bypass
High PSRR
Voltage Regulator
VDD_q
GND
Synthesis Stage
CKIN1P_A
CKOUT1N_A
÷ N31
DSPLL
®
A
CKIN1N_A
÷ N32
CKOUT1P_A
Output Stage
f
OSC
÷ NC1_HS
Input
Monitor
f
3
÷ N2
PLL Bypass
Input Sta ge
CKIN1P_D
÷ N31
DSPLL
®
D
CKIN1N_D
÷ N32
f
OSC
÷ NC1_HS
Input
Monitor
f
3
÷ N2
PLL Bypass
CKIN1P_C
÷ N31
DSPLL
®
C
CKIN1N_C
÷ N32
f
OSC
÷ NC1_HS
Input
Monitor
f
3
÷ N2
PLL Bypass
RSTL_q
CS_q
SCL SDA
LOL_q
IRQ_q
Low Jitter
XO or Clock
OSC_P/N
÷ NC1
PLL Bypass
CKOUT1N_B
÷ NC1
PLL Bypass
CKOUT1P_B
CKOUT1N_C
PLL Bypass
CKOUT1P_C
÷ NC1
CKOUT1N_D
÷ NC1
PLL Bypass
CKOUT1P_D
3.16. Si5375
The Si5375 is a highly integrated , 4- PLL jit ter- atten uat ing precision clock multiplier for applications requiring sub 1
ps jitter performance. Each of the DSPLL® clock multiplier engines accepts an input clock ranging from 2 kHz to
710 MHz and generates an output clock ranging from 2 kHz to 808 MHz. Each DSPLL provides virtually any
frequency translation combination across this operating range. For asynchronous, free-running clock generation
applications, the Si5375’s reference oscillator can be used as a clock source for any of the four DSPLLs. The
Si5375 input clock frequency and clock multiplication ratio are programma ble through an I2C interface. The Si5375
is based on Silicon Laboratories' third-generation DSPLL® technology, which provides any-frequency synthesis
and jitter attenuation in a highly integrated PLL solution that eliminates the need for external VCXO and loop filter
components. Each DSPLL loop bandwidth is digitally programmable from 60 Hz to 8 kHz, providing jitter
performance optimization at the application level. The device operates from a single 1.8 or 2.5 V supply with onchip voltage regulators with excellent PSRR. The Si5375 is ideal for providing clock multiplication and jitter
attenuation in high port count optical line cards requiring independent timing domains.
Figure 15. Si5375 Functional Block Diagram
Rev. 0.531
Si53xx-RM
V
ISE
, V
OSE
VID,V
OD
Differential I/Os
V
ICM
, V
OCM
Single-Ended
Peak-to-Peak Voltage
Differential Peak-to-Peak Voltage
SIGNAL +
SIGNAL –
(SIGNAL +) – (SIGNAL –)
V
t
SIGNAL +
SIGNAL –
VID = (SIG N A L + ) – (S IG NAL–)
V
ICM
, V
OCM
t
F
t
R
80%
20%
CK IN, CKOUT
4. Device Specifications
The following tables are intended to simplify device selection. The specifications in the individual device data
sheets take precedence over this document. Refer to the respective device data sheet for devices not listed in the
tables below.
Si5316
Si5322
1
Min Typ Max Unit
Si5324
Si5325
Si5365
Si5366
Si5367
Si5368
–402585ºC
Table 3. Recommended Operating Conditions
ParameterSymbol Test Condition
Ambient
T
A
Temperature
Supply Voltage
V
DD
3.3 V Nominal
Note 2 Note 2 Note 2 Note 2
2.97 3.3 3.63V
During Normal
Operation
2.5 V Nominal
1.8 V Nominal
Note:
1. All minimum and maximum specifications are guaranteed and apply across the recommended operating conditions.
Typical values apply at nominal supply voltages and an operating temperature of 25 ºC unless otherwise stated.
2. See Sections 6.7.1 and 8.2.1 for restrictions on output formats for TQFP devices at 3.3 V.
2.25 2.5 2.75V
1.71 1.8 1.89V
32Rev. 0.5
Figure 16. Differential Voltage Characteristics
Figure 17. Rise/Fall Time Characteristics
Table 4. DC Characteristics
Si53xx-RM
ParameterSymbolTest Condition
Supply Current
(Independent of
Supply Voltage)
CKIN_n Input Pins
Input Common
Mode Voltage
(Input Threshold
Voltage)
V
I
DD
ICM
LVPECL Format
622.08 MHz Out
All CKOUT’s
LVPECL Format
622.08 MHz Out
Only 1 CKOUT
CMOS Format
19.44 MHz Out
All CKOUTs
CMOS Format
19.44 MHz Out
Only 1 CKOUT
Disable Mode
1
1.8 V ± 10%
2.5 V ± 10%
3.3 V ± 10%
Enabled
Enabled
Enabled
Enabled
Si5316
Si5322
Si5324
Si5325
Si5365
Si5366
Si5367
Si5368
MinT ypMaxUnits
—251279mA
—394435mA
—217243mA
—253284mA
—204234mA
—278321mA
—194220mA
—229261mA
—165— mA
0.9—1.4V
1.0—1.7
1.1—1.95
Input ResistanceCKN
Single-Ended Input
Voltage Swing
Differential Input
Voltage Swing
Notes:
1. Refer to Section 6.7.1 and 8.2.1 for restrictions on output formats for TQFP devices at 3.3 V.
2. This is the amount of leakage that the 3L inputs can tolerate from an external driver. See Figure 55 on page 115.
3. No under- or overshoot is allowed.
RIN
V
ISE
V
ID
Single-ended
f
< 212.5 MHz
CKIN
See Figure 16.
f
>212.5MHz
CKIN
See Figure 16.
f
< 212.5 MHz
CKIN
See Figure 16.
f
>212.5MHz
CKIN
See Figure 16.
204060k
0.2——V
0.25——V
0.2——V
0.25——V
PP
PP
PP
PP
Rev. 0.533
Si53xx-RM
Table 4. DC Characteristics (Continued)
ParameterSymbolTest Condition
Si5316
Si5322
Si5324
Si5325
Si5365
Si5366
Si5367
MinT ypMaxUnits
Si5368
Output Clocks
(CKOUTn—See “8.2. Output Clock Drivers” for Configuring Output Drivers for LVPECL/CML/LVDS/CMOS)
Common ModeV
Differential Output
Swing
Single Ended
Output Swing
Differential Output
CKO
Voltage
Common Mode
CKO
Output Voltage
Differential
CKO
Output Voltage
OCM
V
OD
V
SE
VCM
LVPECL 100
load line-to-line
LVPECL 100
load line-to-line
LVPECL 100
load line-to-line
CML 100 load
VD
line-to-line
CML 100 load
line-to-line
LVDS 100 load
VD
line-to-line
Low swing LVDS
1
1
VDD–
1.42
—VDD–
1.25
1.1—1.9V
0.5—0.93V
350425500mV
—VDD
—V
– .36
500700900mV
350425500mV
V
PP
PP
PP
PP
PP
100 load
line-to-line
Common Mode
Output Voltage
CKO
LVDS 100 load
VCM
line-to-line
1.1251.2 1.275V
Differential Output
Resistance
CKO
RD
CML, LVPECL,
LVDS, Disabled,
170200230
Sleep
Output Volt age LowCKO-
VOLLH
Output Volt age HighCKO-
VOHLH
Notes:
1. Refer to Section 6.7.1 and 8.2.1 for restrictions on output formats for TQFP devices at 3.3 V.
2. This is the amount of leakage that the 3L inputs can tolerate from an external driver. See Figure 55 on page 115.
1. Refer to Section 6.7.1 and 8.2.1 for restrictions on output formats for TQFP devices at 3.3 V.
2. This is the amount of leakage that the 3L inputs can tolerate from an external driver. See Figure 55 on page 115.
3. No under- or overshoot is allowed.
ILL
IMM
IHH
ILL
See note
2
——0.15 x
0.45 x
V
0.85 x
V
—0.55x
DD
—— V
DD
–20——µA
V
V
DD
V
V
DD
Rev. 0.535
Si53xx-RM
Table 4. DC Characteristics (Continued)
ParameterSymbolTest Condition
Input Mid CurrentI
Input High CurrentI
IMM
IHH
See note
See note
2
2
Si5316
Si5322
Si5324
Si5325
Si5365
Si5366
Si5367
MinT ypMaxUnits
Si5368
–2—2µA
——20µA
LVCMOS Output Pins
Output Volt age LowV
Output Volt age HighV
Tri-State Leakage
I
OL
OH
OZ
IO=2mA
V
=1.62V
DD
I
=2mA
O
V
=2.97V
DD
IO=–2mA
V
=1.62V
DD
I
=–2mA
O
V
=2.97V
DD
RST =0
——0.4 V
——0.4 V
V
DD –
—— V
0.4
V
DD –
—— V
0.4
–100—100µA
Current
Notes:
1. Refer to Section 6.7.1 and 8.2.1 for restrictions on output formats for TQFP devices at 3.3 V.
2. This is the amount of leakage that the 3L inputs can tolerate from an external driver. See Figure 55 on page 115.
3. No under- or overshoot is allowed.
36Rev. 0.5
Si53xx-RM
Table 5. DC Characteristics—Microprocessor Devices (Si5324, Si5325, Si5367, Si5368)
ParameterSymbolTest ConditionMinTypMaxUnits
2
C Bus Lines (SDA, SCL)
I
Input Voltage LowV
Input Voltage HighV
Input CurrentI
Hysteresis of Schmitt trig-
V
ger inputs
Output Volt age LowV
ILI2C
IHI2C
II2C
HYSI2C
OHI2C
VIN = 0.1 x V
to 0.9 x V
DD
DD
VDD = 1.8 V0.1 x V
V
= 2.5 or 3.3 V0.05 x V
DD
VDD=1.8V
——0.25 x V
0.7 x V
DD
—V
DD
DD
–10—10µA
DD
—— V
—— V
DD
——0.2xVDDV
V
V
IO = 3 mA
= 2.5 or 3.3 V
V
DD
——0.4 V
IO = 3 mA
Table 6. SPI Specifications (Si5324, Si5325, Si5367, and Si5368)
ParameterSymbolTest ConditionsMinTypMaxUnit
Duty Cycle, SCLKt
Cycle Time, SCLKt
Rise Time, SCLKt
Fall Time, SCLKt
Low Time, SCLKt
High Time, SCLKt
Delay Time, SCLK Fall to SDO Activet
Delay Time, SCLK Fall to SDO Transitiont
Delay Time, SS Rise to SDO Tri-state t
Setup Time, SS to SCLK Fallt
Hold Time, SS to SCLK Riset
Setup Time, SDI to SCLK Riset
Hold Time, SDI to SCLK Riset
Delay Time between Slave Selectst
Note: All timing is referenced to the 50% level of the waveform unless otherwise noted. Input test levels are VIH = VDD–4V,
2. BWSEL [1:0] loop bandwidth settings provided in Pin Descriptions.
GEN
= f
IN
= 622.08 MHz, LVPECL clock input: 1.19 Vppd with 0.5 ns rise/fall time (20–80%), LVPECL
OUT
0.012–201096—.48—ps
1,2
DSPLL
Bandwidth
(kHz)
2
MinTypMaxUnit
rms
rms
rms
rms
44Rev. 0.5
Si53xx-RM
Table 11. Thermal Characteristics
ParameterS ymbolTest ConditionDevicesValueUnit
Thermal Resistance Junction to Ambient
Thermal Resistance Junction to Case
JA
JC
Still AirSi5316, Si5319,
Si5322, Si5323,
Si5324, Si5325
Si5365, Si5366,
Si5367, Si5368
Still AirSi5316, Si5319,
Si5322, Si5323,
Si5324, Si5325
32ºC/W
40ºC/W
14ºC/W
Rev. 0.545
Si53xx-RM
f
IN
DSPLL
Phase
Detector
Digital
DCO
Digital Loop
Filter
Fvco
M
f
OUT
5. DSPLL (All Devices)
All members of the Any-Frequency Precision Clocks family incorporate a phase-locked loop (PLL) that utilizes
Silicon Laboratories' third generation DSPLL technology to eliminate jitter, noise, and the need for external VCXO
and loop filter components found in discrete PLL implementations. This is achieved by using a digital signal
processing (DSP) algorithm to replac e the loop filter commonly found in discrete PLL designs. Because external
PLL components are not required, sensitivity to board-level noise sources is minimized. This digital technology
provides highly stable and consistent operation over process, temperature, and voltage variations.
A simplified block diagram of the DSPLL is shown in Figure 20. This algorithm processes the phase detector error
term and generates a digital frequency control word M to adjust the frequency of the digitally-controlled oscillator
(DCO). The narrowband configuration devices (Si5316, Si5319, Si5323, Si5324, Si5326, Si5327, Si5366, Si5368,
and Si5369) provide ultra-low jitter generation by using an external jitter reference clock and jitter attenuation. For
applications where basic frequency multiplication of low jitter clocks is all that is required, the wideband parts
(Si5322, Si5325, Si5365, and Si5367) are available.
Fundamental to these parts is a clock multiplication circuit th at is simplified in Figure 21. By having a large range of
dividers and multipliers, nearly any output frequency can be created from a fixed input frequency. For typical
telecommunications and data communications applications, the hardware control parts (Si5316, Si5322, Si5323,
Si5365, and Si5366) provide simple pin control.
The microprocessor controlled parts (Si5319, Si5324, Si5325, Si5326, Si5327, Si5367, Si5368, and Si5369)
provide a programmable range of clock multiplications. To assist users in finding valid divider settings for a
particular input frequency and clock multiplication ratio, Silicon Laboratories offers PC-based software (DSPLLsim)
that calculates these settings automatically. When multiple divider combinations produce the same output
frequency, the software recommends the divider settings yielding the recommended settings for phase noise
performance and power consumption.
Figure 21. Clock Multiplication Circuit
Rev. 0.547
Si53xx-RM
Jitter
Transfer
0 dB
BW
f
Jitter
Peaking
–20 dB/dec.
Jitter Out
Jitter In
)
(
20 x LOG
5.2. PLL Performance
All members of the Any-Frequency Precision Clock family of devices provide extremely low jitter generation, a wellcontrolled jitter transfer function, and high jitter tolerance. For mor e information the loop bandwidth and its effect on
jitter attenuation, see "Appendix H—Jitter Attenuation and Loop BW" on page 164.
5.2.1. Jitter Generation
Jitter generation is defined as the amount of jitter produced at the output of the device with a jitter free input clock.
Generated jitter arises from sources within the VCO and other PLL components. Jitter generation is a function of
the PLL bandwidth setting. Higher loop bandwidth settings may result in lower jitter generation, but may result in
less attenuation of jitter that might be present on the input clock signal.
5.2.2. Jitter Transfer
Jitter transfer is defined as the ratio of output signal jitter to input signal jitter for a specified jitter frequency. The
jitter transfer characteristic determines the amount of input clock jitter that passes to the outputs. The DSPLL
technology used in the Any-Frequency Precision Clock devices provides tightly controlled jitter transfer curves
because the PLL gain parameters are determined larg ely by digital circuits which do not vary over supply voltage,
process, and temperature. In a system application, a well-controlled transfer curve minim izes the output clock jitter
variation from board to board and provides more consistent system level jitter performance.
The jitter transfer characteristic is a function of the loop bandwidth setting. Lower bandwidth settings result in more
jitter attenuation of the incoming clock, but may result in higher jitter generation. Section 1 Any-Frequency
Precision Clock Product Family Overview also includes specifications related to jitter bandwidth and peaking.
Figure 22 shows the jitter transfer curve mask.
Figure 22. PLL Jitter Transfer Mask/Template
48Rev. 0.5
Si53xx-RM
Input
Jitter
Amplitude
A
j0
–20 dB/dec.
f
Jitter In
Excessive Input Jitter Range
BW/100 BW/10
BW
A
j0
5000
BW
------------ -
ns pk-pk=
A
j0
5000
100
------------ -
50 ns pk-pk==
5.2.3. Jitter Tolerance
Jitter tolerance is defined as the maximum peak-to-peak sinusoidal jitter that can be present on the incoming clock
before the DSPLL loses lock. The tolerance is a function of the jitter frequency, because tolerance improves for
lower input jitter frequency.
The jitter tolerance of the DSPLL is a function of the loop bandwidth setting. Figure 23 shows the general shape of
the jitter tolerance curve versus input jitter frequency. For jitter frequencies above the loop bandwid th, the tolerance
is a constant value A
lower input jitter frequencies.
. Beginning at the PLL bandwidth, the tolerance increases at a rate of 20 dB/decade for
j0
The equation for the high frequency jitter tolerance can be expressed as a function of the PLL loop bandwidth (i.e.,
bandwidth):
For example, the jitter tolerance when f
Figure 23. Jitter Tolera nce Mask/Template
= 155.52 MHz, f
in
= 622.08 MHz and the loop bandwidth (BW) is 100 Hz:
out
Rev. 0.549
Si53xx-RM
6. Pin Control Parts (Si5316, Si5322, Si5323, Si5365, Si5366)
These parts provide high-performance clock multiplication with simple pin control. Many of the control inputs are
three levels: High, Low, and Medium. High and Low are standard voltage levels determined by the supply voltage:
and Ground. If the input pin is left floating, it is driven to nominally half of VDD. Effectively, this creates three
V
DD
logic levels for these controls.
These parts span a range of applications and I/O capacity as shown in Table 12.
Table 12. Si5316, Si5322, Si5323, Si5365 and Si5366 Key Features
Si5316Si5322Si5323Si5365Si5366
SONET Frequencies
DATACOM Frequencies
DATACOM/SONET internetworking
Fixed Ratio between input clocks
Flexible Frequency Plan
Number of Inputs22244
By setting the tri-level FRQSEL[3:0] pins these devices provide a wide range of standard SONET and data
communications frequency scaling, including simple integer frequency multiplication to fractional settings required
for coding and decoding.
6.1.1. Clock Multiplication (Si5316 )
The device accepts dual input clocks in the 19, 39, 78, 155, 311, or 622 MHz frequency range and generates a dejittered output clock at the same frequency. The frequency range is set by the FRQSEL [1:0] pins, as shown in
Table 13.
Table 13. Frequency Settings
FRQSEL[1:0]Output Frequency (MHz)
LL19.38–22.28
LM38.75–44.56
LH77.50–89.13
ML155.00–178.25
MM310.00–356.50
MH620.00–710.00
50Rev. 0.5
Si53xx-RM
1, 4, 32
1, 4, 32
CKIN1
CKIN2
DSPLL
F
out
f3 = F
out
f
3
One-to-one
frequency ratio
The Si5316 can accept a CKIN1 input at a different frequency than the CKIN2 input. The frequency of one input
clock can be 1x, 4x, or 32x the frequency of the other inpu t clock. The o utput frequency is always equal to the lower
of the two clock inputs and is set via the FRQSEL [1:0] pins. The frequency applied at each clock input is divided
down by a pre-divider as shown in the Figure 1 on page 17. These pre-dividers must be set such that the two
resulting clock frequencies, f3_1 and f3_2 must be equal and are set by the FRQSEL [1:0] pins. Input divider
settings are controlled by the CK1DIV and CK2DIV pins, as shown in Table 14.
6.1.2. Clock Multiplicatio n (Si5322, Si5323, Si5365, Si5366)
These parts provide flexible frequency plans for SONET, DATACOM, and interworking be tween the two (Table 16,
Table 17, andTa ble 18 respectively). The CKINn inputs must be the same frequency as specified in the tables. Th e
outputs are the same frequency; however, in the Si5365 and Si5366, CKOUT3 and CKOUT4 can be further divided
down by using the DIV34 [1:0] pins.
The following notes apply to Tables 16, 17, and 18:
1. All entries are available for the Si5323 and Si5366. Only those marked entries under the WB column are
available for the Si5322 and Si5365.
2. The listed output frequencies appear on CKOUTn. For the Si5365 and Si5366, sub- multiples are available on
CKOUT3 and CKOUT4 using the DIV34[1:0] control pins.
3. All ratios are exact, but the frequency values ar e ro u nd ed .
4. For bandwidth settings, f3 values, and frequency operating ranges, consult DSPLLsim.
5. For the Si5366 with CK_CONF = 1, CKIN3 and CKIN4 are the same frequency as FS_OUT.
0LLLL
1LLLM
2LLLH
3LLML
4LLMM
5LLMH
6LLHL25/4 x 66/64161.13
7LLHM51/8 x 66/641 64.36
8LLHH25/4 x 66/64 x 255/238172.64
9LMLL25/4 x 66/64 x 255/237173.37
10LMLM51/8 x 66/64 x 255/238176.1
11LMLH51/8 x 66/64 x 255/237176.84
12LMML
13LMMM
14LMMH25 x 66/64644.53
15LMHL51/2 x 66/64657.42
16LMHM25 x 66/64 x 255/238690.57
17LMHH25 x 66/64 x 255/237693.48
18LHLL51/2 x 66/64 x 255/238704.38
19LHLM51/2 x 66/64 x 255/237707.35
20LHLH
21LHML
22LHMM
23LHMH
24LHHL
25LHHM
26LHHH
27MLLL3/2 x 66/64 x 255/238176.1
15LMHL15625/1944156.25
16LMHM31875/3888159.375
17LMHH15625/1944 x 66/64161.13
18LHLL31875/3888 x 66/64164.36
19LHLM15625/1944 x 66/
64 x 255/238
20LHLH31875/3888 x 66/
64 x 255/238
21LHML10625/972212.5
22LHMM10625/486425
23LHMH15625/486 x 66/64644.53
24LHHL31875/972 x 66/64657.42
25LHHM15625/486 x 66/
64 x 255/238
26LHHH31875/972 x 66/
64 x 255/238
27MLLL27.000127
28MLLM250/9174.17582
172.64
176.1
690.57
704.38
29MLLH11/474.25
Rev. 0.561
Si53xx-RM
Table 18. SONET to Datacom Clock Multiplication Settings (Continued)
Setting FRQSEL[3:0]
30MLML
31MLMM
32MLMH74.17691/25027
33MLHL174.17582
34MLHM91 x 11/250 x 474.25
35MLHH74.2504/1127
36MMLL4 x 250/11 x 9174.17582
37MMLM174.25
38MMLH77.76010625/7776106.25
39MMML3125/1944125
40MMMM15625/7776156.25
41MMMH31875/15552159.375
42MMHL15625/7776 x 66/64161.13
43MMHM31875/15552 x 66/64164.36
44MMHH15625/7776 x 66/
fIN (MHz)Mult Factorf
WB
62.5002125
4250
64 x 255/238
OUT
172.64
* (MHz)
45MHLL31875/15552 x 66/
64 x 255/238
46MHLM10625/3888212.5
47MHLH10625/1944425
48MHML15625/1944 x 66/64644.53
49MHMM31875/3888 x 66/64657.42
50MHMH15625/1944 x 66/
64 x 255/238
51MHHL31875/3888 x 66/
64 x 255/238
176.1
690.57
704.38
62Rev. 0.5
Si53xx-RM
Table 18. SONET to Datacom Clock Multiplication Settings (Continued)
Setting FRQSEL[3:0]
52MHHM155.52015625/15552156.25
53MHHH31875/31104159.375
54HLLL15625/15552 x 66/64161.13
55HLLM31875/31104 x 66/64164.36
56HLLH15625/15552 x 66/
57HLML31875/31104 x 66/
58HLMM10625/7776212.5
59HLMH10625/3888425
60HLHL15625/3888 x 66/64644.53
61HLHM31875/7776 x 66/64657.42
62HLHH15625/3888 x 66/
63HMLL31875/7776 x 66/
fIN (MHz)Mult Factorf
WB
64 x 255/238
64 x 255/238
64 x 255/238
64 x 255/238
OUT
172.64
176.1
690.57
704.38
* (MHz)
64HMLM622.08015625/15552 x 66/64644.53
65HMLH31875/31104x 66/64657.42
66HMML15625/15552 x 66/
64 x 255/238
67HMMM31875/31104 x 66/
64 x 255/238
690.57
704.38
Rev. 0.563
Si53xx-RM
6.1.3. CKOUT3 and CKOUT4 (Si5365 and Si5366)
Submultiples of the output frequency on CKOUT1 and CKOUT2 can be produced on the CKOUT3 and CKOUT4
outputs using the DIV34 [1:0] control pins as shown in Table 19.
The loop bandwidth (BW) is digitally programmable using the BWSEL [1:0] input pins. The device operating
frequency should be determined prior to loop bandwidth configuration because the loop bandwidth is a function of
the phase detector input frequency and the PLL feedback divider setting. Use DSPLLsim to calculate these values
automatically. This utility is available for download from www.silabs.com/timing.
The DCO uses the reference clock on the XA/XB pins as its reference for jitter attenuation . The XA/XB pins support
either a crystal oscillator or an input buffer (single-ended or differential) so that an external oscillator can be used
as the reference source. The reference source is chosen with the RATE [1:0] pins. In both cases, there are wide
margins in the absolute frequency of the reference input because it is a fixed frequency reference and is only used
as a jitter reference and holdover reference (see "6.4. Digital Hold/VCO Freeze" on page 70).
However, care must be taken in certain areas for optimum performance. For details on this subject, refer to
"Appendix B—Frequency Plans and Jitter Performance (Si5316, Si5319, Si5323, Si5324, Si5326, Si5327, Si5366,
Si5368, Si5369, Si5374, Si5375)" on page 121. For examples of connections to the XA/XB pins, refer to "8.4.
Crystal/Reference Clock Interfaces (Si5316, Si5319, Si5323, Si5324, Si5326, Si5327, Si5366, Si5368, Si5369,
Si5374, and Si5375)" on page 113.
The input-to-output skew for these devices is not controlled.
6.1.8. Wideband Performance (Si5322 and Si5365)
These devices operate as wideband clock multipliers without an external resonator or reference clock. They are
ideal for applications where the input clock is already low jitter and only simple clock multiplication is required. A
limited selection of clock multiplication factors is available (See Table 16, Table 17, andTable 18).
6.1.9. Lock Detect (Si5322 and Si5365)
A PLL loss of lock indicator is not available in these parts.
6.1.10. Input-to-Output Skew (Si5322 and Si5365)
The input-to-output skew for these devices is not controlled.
64Rev. 0.5
Si53xx-RM
6.2. PLL Self-Calibration
An internal self-calibration (ICAL) is performed before operation to optimize loop parameters and jitter
performance. While the self-calibration is being performed, the DSPLL is being internally controlled by the selfcalibration state machine, and the LOL alarm will be active for narrowband parts. The self-calibration time t
is given in Table 8, “AC Characteristics—All Devices”.
Any of the following events will trigger a self-calibration:
Power-on-reset (POR)
Release of the external reset pin RST (transition of RST from 0 to 1)
Change in FRQSEL, FRQTBL, BWSEL, or RATE pins
Internal DSPLL registers out-of-range, indicating the need to relock the DSPLL.
In any of the above cases, an internal self-calibration will be initiated if a valid input clock exists (no input alarm)
and is selected as the active clock at that time. For the Si5316, Si5323 and Si5366, the external crystal or
reference clock must also be present for the self-calibration to begin. If valid clocks are not present, the selfcalibration state machine will wait until they appear, at which time the calibration will start. All outputs are on during
the calibration process.
After a successful self-calibration has been performed with a valid input clock, no subsequent self-calibrations are
performed unless one of the above conditions are met. If the input clock is lost following self-calibration, the device
enters digital hold mode. When the input clock returns, the device relocks to the input clock without performing a
self-calibration. (Narrow band devices only).
6.2.1. Input Clock St ability during Internal Self-Calibration (Si5316, Si5322, Si5323, Si5365, Si5366)
An exit from reset must occur when the selected CKINn clock is stable in frequency with a frequency value that is
within the operating range that is reported by DSPLLsim. The other CKINs must also either be stable in frequency
or squelched during a reset.
6.2.2. Self-Calibration caused by Changes in Input Frequency (Si5316, Si5322, Si5323, Si5365, Si5366)
If the selected CKINn varies by 500 ppm or more in frequency since the last calibration, the device may initiate a
self-calibration.
Follow the recommended RESET guidelines in Table 20 and Table 21 when reset should be applied to a device.
LOCKHW
Rev. 0.565
Si53xx-RM
Table 20. Si5316, Si5322, and Si5323 Pins and Reset
Pin #Si5316 Pin
Name
2N/AFRQTBLFRQTBLYes
11RATE 0N/ARATE 0Yes
14DBL_BYDBL2_BYDBL2_BYNo
15RATE1N/ARATE1Yes
19N/AN/ADECNo
20N/AN/AINCNo
22BWSEL0BWSEL0BWSEL0Yes
23BWSEL1BWSEL1BWSEL1Yes
24FRQSEL0FRQSEL0F RQSEL0Yes
25FRQSEL1FRQSEL1F RQSEL1Yes
26N/AFRQSEL2FRQSEL2Yes
27N/AFRQSEL3FRQSEL3Yes
30SFOUT1N/ASFOUT1No, but skew not guaranteed without Reset
33SFOUT0N/ASFOUT0No, but skew not guaranteed without Reset
Si5322 Pin
Name
Si5323 Pin
Name
Must Reset after Changing
Table 21. Si5365 and Si5366 Pins and Reset
Pin #Si5365 Pin NameSi5366 Pin NameMust Reset after Changing
4FRQTBLFRQTBLYes
32N/ARATE 0Yes
42N/ARATE 1Yes
51N/ACK_CONFYes
54N/ADECNo
55N/AINCNo
60BWSEL0BSWEL0Yes
61BWSEL1BWSEL1Yes
66DIV34_0DIV34_0Yes
67DIV34_1DIV34_1Yes
68FRQSEL0FRQSEL0Yes
69FRQSEL1FRQSEL1Yes
70FRQSEL2FRQSEL2Yes
71FRQSEL3FRQSEL3Yes
80N/ASFOUT1No, but skew not guaranteed without Reset
95N/ASFOUT0No, but skew not guaranteed without Reset
66Rev. 0.5
Si53xx-RM
6.3. Pin Control Input Clock Control
This section describes the clock selection capabilities (manual input selection, automatic input selection, hitless
switching, and revertive switching). When switching between two clocks, LOL may temporarily go high if the two
clocks differ in frequency by more than 100 ppm.
6.3.1. Manual Clock Selection
Manual control of input clock selection is chosen via the CS[1:0] pins according to Table 22 and Table 23.
The manual input clock selection settings for the Si5365 and the Si5366 are shown in Table 23. The Si5366 has
two modes of operation (See Section “6.5. Frame Synchronization (Si5366)”). With CK_CONF = 0, any of the four
input clocks may be selected manually; however, when CK_CONF = 1 th e inputs are paired, CKIN1 is paired with
CKIN3 and likewise for CKIN2 and CKIN4. Therefore, only two settings are available to select on e of the two pairs.
The AUTOSEL input pin sets the input clock selection mode as shown in Table 24. Automatic switching is either
revertive or non-revertive. Setting AUTOSEL to M or H, changes the CSn_CAm pins to output pins that indicate the
state of the automatic clock selection (See Table 25 and Table 26). Digital hold is indicated by all CnB signals going
high after a valid ICAL.
Table 24. Automatic/Manual Clock Selection
AUTOSELClock Selection Mode
LManual (See Previous Section)
MAutomatic Non-revertive
HAutomatic Revertive
Table 25. Clock Active Indicators (AUTOSEL = M or H) (Si5322 and Si5323)
CS_CAActive Clock
0CKIN1
1CKIN2
Table 26. Clock Active Indicators (AUTOSEL = M or H) (Si5365 and Si5367)
CA1CA2CS0_CA3CS1_CA4Active Clock
1000CKIN1
0100CKIN2
0010CKIN3
0001CKIN4
The prioritization of clock inputs for automatic switching is shown in Table 27 and Table 28. This priority is
hardwired in the devices.
Table 27. Input Clock Priority for Auto Switching (Si5322, Si5323)
PriorityInput Clocks
1CKIN1
2CKIN2
3Digital Hold
68Rev. 0.5
Si53xx-RM
Table 28. Input Clock Priority for Auto Switching (Si5365, Si5366)
PriorityInput Clock Configuration
Si5365Si5366
4 Input Clocks
(CK_CONF = 0)
1CKIN1CKIN1/CKIN3
2CKIN2CKIN2/CKIN4
3CKIN3N/A
4CKIN4N/A
5Digital HoldDigital Hold
At power-on or reset, the valid CKINn with the highest priority (1 being the highest priority) is automatically
selected. If no valid CKINn is available, the device suppresses the output clocks and waits for a valid CKINn signal.
If the currently selected CKINn goes into an alarm state, the next valid CKINn in priority order is selected. If no valid
CKINn is available, the device enters Digital Hold.
Operation in revertive and non- revertive is different when a signal becomes valid:
Revertive (AUTOSEL = H):The device cons tantly monitors all CKINn. If a CKINn with a higher priority than
the current active CKINn becomes valid, the active CKINn is changed to the
CKINn with the highest priority.
Non-revertive (AUTOSEL = M): The active clock does not change until there is an alarm on the active clock. The
device will then select the highest priority CKINn that is valid. Once in digital hold,
the device will switch to the first CKINn that becomes valid.
6.3.3. Hitless Switching with Phase Build-Out (Si5323, Si5366)
Silicon Laboratories switching technology performs “phase build-out” to minimize the propagation of phase
transients to the clock outputs during input clock switching. All switching between input clocks occurs within the
input multiplexor and phase detector circuitry. The phase detector circuitry continually monitors the phase
difference between each input clock and the DSPLL output clock, f
clock signal at a specified phase offset relative to f
At the time a clock switch occurs, the phase detector circu itr y kn ows bo th th e inpu t- to -o utpu t p ha se re lation sh ip for
the original input clock and for the new input clock. The phase detector circuitry locks to the new input clock at the
new clock's phase offset so that the phase of the output clock is not disturbed. The phase difference between the
two input clocks is absorbed in the phase detector's offset value, rather than being propagated to the clock output.
The switching technology virtually eliminates the output clock phase transients traditionally associated with clock
rearrangement (input clock switching).
so that the phase offset is maintained by the PLL circuitry.
OSC
FSYNC Switching
(CK_CONF = 1)
. The phase detector circuitry can lock to a
OSC
Rev. 0.569
Si53xx-RM
6.4. Digital Hold/VCO Freeze
All Any-Frequency Precision Clock devices feature a hold over or VCO freeze mo de, where by the DSPL L is locked
to a digital value.
6.4.1. Narrowband Digital Hold (Si5316, Si5323, Si5366)
If an LOS or FOS condition exists on the selected input clock, the device enters digital hold. In this mode, the
device provides a stable output frequency until the input clock returns and is validated. When the device enters
digital hold, the internal oscillator is initially held to its last frequency value. Next, the internal oscillator slowly
transitions to a historical average frequency value that was taken over a time window of 6,711 ms in size that
ended 26 ms before the device entered digit al hold. This frequen cy value is t aken from an internal memory locatio n
that keeps a record of previous DSPLL frequency values. By using a historical average frequency, input clock
phase and frequency transients that may occur immediately preceding loss of clock or any event causing digital
hold do not affect the digital hold frequency. Also, noise related to input clock jitter or internal PLL jitter is
minimized.
If a highly stable reference, such as an oven-controlled crystal oscillator, is supplied at XA/XB, an extremely stable
digital hold can be achieved. If a crystal is supplied at the XA/XB port, the digital hold stability will be limited by the
stability of the crystal.
6.4.2. Recovery from Digital Hold (Si5316 , Si5323, Si5366)
When the input clock signal returns, the device transitions from digital hold to the selected input clock. The device
performs hitless recovery from digital hold. The clock transition from digital hold to the returned input clock includ es
“phase buildout” to absorb the phase difference between the digital hold clock phase and the input clock phase.
6.4.3. Wideband VCO Freeze (Si5322, Si5365)
If an LOS condition exists on the selected input clock, the device freezes the VCO. In this mode, the device
provides a stable output frequency until the input clock returns and is validated. When the device enters VCO
freeze, the internal oscillator is initially held to its last frequency value.
6.5. Frame Synchronization (Si5366)
FSYNC is used in applications that require a synchronizing pulse that has an exact number of periods of a highrate clock, Frame Synchronization is selected by setting CK_CONF = 1 and FRQTBL = L). In a typical frame
synchronization application, CKIN1 and CKIN2 are high-speed input clocks from primary and secondary clock
generation cards and CKIN3 and CKIN4 are their associated primary and secondary frame synchronization
signals. The device generates four output clocks and a frame sync output FS_OUT. CKIN3 and CKIN4 control the
phase of FS_OUT.
The frame sync inputs supplied to CKIN3 and CKIN4 must be 8 kHz. Since the frequency of FS_OUT is derived
from CKOUT2, CKOUT2 must be a standard SONET frequency (e.g. 19.44 MHz, 77.76 MHz). Table 16 lists the
input frequency/clock multiplication ratio combinations supporting an 8 kHz output on FS_OUT.
70Rev. 0.5
Si53xx-RM
6.6. Output Phase Adjust (Si5323, Si5366)
Overall device skew (CKINn to CKOUT_n phase delay) is controllable via the INC and DEC input pins . A positive
pulse applied at the INC pin increases the device skew by 1/f
the DEC pin decreases the skew by the same amount. Since f
control is approximately 200 ps. Using the INC and DEC pins, there is no limit to the range of skew adjustment that
can be made. Following a power-up or reset, the skew will revert to the reset value.
The INC pin function is not available for all frequency table selections. DSPLLsim reports this whenever it is used
to implement a frequency plan.
6.6.1. FSYNC Realignment (Si5366)
The FS_ALIGN pin controls the realignment of FS_OUT to the active CKIN3 or CKIN4 input. The currently active
frame sync input is determined by which input clock is currently being used by the PLL. For example, if CKIN1 is
being selected as the PLL inp ut, CKIN3 is the currently-active frame sync input. If neither CKIN3 or CKIN4 are
currently active (digital hold), the realignment request is ignored. The active edge used for realignment is the
CKIN3 or CKIN4 rising edge.
FS_ALIGN operates in Level Sensitive mode (See Figure 19, “Frame Synchronization Timing.”). While FS_ALIGN
is active, each active edge of the curren tly-active frame sync input (CKIN3 or CKIN 4) is used to control the NC5
output divider and therefore the FS_OUT phase. Note that while the realignment control is active, it cannot be
guaranteed that a fixed number of high-frequency clock (CKOUT2) cycles exists between each FS_OUT cycle.
The resolution of the phase realignment is 1 clock cycle of CKOUT2. If the realignment control is not active, the
NC5 divider will continuously divide down its f
clock (CKOUT2) cycles between each FS_OUT cycle.
At power-up or any time after the PLL has lost lock and relo cked, th e device a utomatica lly perform s a realign ment
of FS_OUT using the currently active sync input. After this, as long as the PLL remains in lock and a realignment is
not requested, FS_OUT will include a fixed number of high-speed clock cycles, even if input clock switches are
performed. If many clock switches are performed in phase build-out mode, it is possible that the input sync to
output sync phase relationship will shift due to the accumulated residual phase transients of the phase build-out
circuitry. If the sync alignment error exceeds the threshold in either the positive or negative direction, an alignment
alarm becomes active. If it is then desired to reestablish the desired input-to-output sync phase relationship, a
realignment can be performed. A realignment request may cause FS_OUT to instantaneously shift its output edge
location in order to align with the active input sync phase.
6.6.2. Including FSYNC Inputs in Clock Selection (Si5366)
The frame sync inputs, CKIN3 and CKIN4, are both monitored for loss-of-signal (LOS3_INT and LOS4_INT)
conditions. To include these LOS alarms in the input clock selection algorithm, set FS _SW = 1. The LOS3_INT is
logically ORed with LOS1_INT and LOS4_INT is ORed with LOS2_INT as inputs to the clock selection state
machine. If it is desired not to include these alarms in the clock selection algorithm, set FS_SW = 0. The FOS
alarms for CKIN3 and CKIN4 are ignored. See Table 33 on page 74.
6.6.3. FS_OUT Polarity and Pulse Width Control (Si5366)
Additional output controls are available for FS_OUT. FS_OUT is active high, and the pulse width is equal to one
period of the CKOUT2 output clock. For example, if CKOUT2 is 622.08 MHz, the FS_OUT pulse width will be 1/
622.08e6 = 1.61 ns.
6.6.4. Using FS_OUT as a Fifth Output Clock (Si5366)
In applications where the frame synchronization functionality is not needed, FS_OUT can be used as a fifth clock
output. In this case, no realignment requests should be made to the NC5 divider. (This is done by holding
FS_ALIGN to 0 and CK_CONF = 0).
CKOUT2
input. This guarantees a fixed number of high-frequency
, one period of the DCO output clock. A pulse on
OSC
is close to 5 GHz, the resolution of the skew
OSC
Rev. 0.571
Si53xx-RM
6.6.5. Disabling FS_OUT (Si5366)
The FS_OUT maybe disabled via the DBLFS pin, see Table 29. The additional state (M) provided allows for
FS_OUT to drive a CMOS load while the other clock outputs use a different signal format as specified by the
SFOUT[1:0] pins.
Table 29. FS_OUT Disable Control (DBLFS)
DBLFSFS_OUT State
HTri-State/Powerdown
MActive/CMOS Format
LActive/SFOUT[1:0] Format
6.7. Output Clock Drivers
The devices include a flexible output driver structure that can drive a variety of loads, including LVPECL, LVDS,
CML, and CMOS formats. The signal format is selec ted jointly for all outputs using the SFOUT [1:0] pins, which
modify the output common mode and differential signal swing. See Table 4, “DC Characteristics” for output driver
specifications. The SFOUT [1:0] p ins are three- level input pins, with the states designated as L (ground), M (V
2), and H (V
Table 30 shows the signal formats based on the supply voltage and the type of load being driven. For the CMOS
setting (SFOUT = LH), both output pins drive single-ended in-phase signals and should be externally shorted
together to obtain the drive strength specified in Table 4, “DC Characteristics”, see Section“8.2. Output Clock
Drivers”.
DD
).
DD
/
Table 30. Output Signal Format Selection (SFOUT)
SFOUT[1:0]Signal Format
HLCML
HMLVDS
LHCMOS
LMDisabled
MHLVPECL
MLLow-swing LVDS
All OthersReserved
The SFOUT [1:0] pins can also be used to disable the output. Disabling th e output p ut s the CKOUT+ and CKOUT–
pins in a high-impedance state relative to V
each other through a 200 on-chip resistance (differential impedance of 200 ). The maximum amount of internal
circuitry is powered down, minimizing power consumption and noise generation. Changing SFOUT without a reset
causes the output to output skew to become random. When SFOUT = LH for CMOS, PLL bypass mode is not
supported.
6.7.1. LVPECL and CMOS TQFP Output Signal Format Restrictions at 3.3 V (Si5365, Si5366)
The LVPECL and CMOS output formats draw more current than either LVDS or CML. However, the allowed ou tp ut
format pin settings are restricted so that the ma xi mum p ower dissipation for the TQFP devices is limited when they
are operated at 3.3 V. When SFOUT[1:0] = MH or LH (for either LVPECL or CMOS), either DBL5 must be H or
DBL34 must be high.
(common mode tri-state) while the two outputs remain connected to
DD
72Rev. 0.5
Si53xx-RM
6.8. PLL Bypass Mode
The device supports a PLL bypass mode in which the selected input clock is fed directly to all enabled output
buffers, bypassing the DSPLL. In PLL bypass mode, the input and output clocks will be at the same frequency. PLL
bypass mode is useful in a laboratory environment to measure system performance with and without the effects of
jitter attenuation provided by the DSPLL.
The DSBL2/BYPASS pin is used to select the PLL bypass mode according to Table 31.
Table 31. DSBL2/BYPASS Pin Settings
DSBL2/BYPASSFunction
LCKOUT2 Enabled
MCKOUT2 Disabled
HPLL Bypass Mode w/ CKOUT2 Enabled
Internally, the bypass path is implemented with high-speed differential signaling for low jitter. Bypass mode does
not support CMOS clock output.
6.9. Alarms
Summary alarms are available to indicate the overall status of the input signals and fr ame a lignment (Si53 66 o nly).
Alarm outputs stay high until a ll the alarm conditions for that alarm output are cleared.
The device has loss-of-signal circuitry that continuously monitors CKINn for missing pulses. The LOS circuitry
generates an internal LOSn_INT output signal that is processed with other alarms to generate CnB.
An LOS condition on CKIN1 causes the internal LOS1_INT alarm to become ac tive. Similarly, an LOS condition on
CKINn causes the LOSn_INT alarm to become active. Once a LOSn_INT alarm is asserted on one of the input
clocks, it remains asserted until that input clock is validated over a designated time period. The time to clear
LOSn_INT after a valid input clock appears is listed in Table 8, “AC Characteristics—All Devices”. If another error
condition on the same input clock is detected during the validation time then the alarm remains asserted and the
validation time starts over.
6.9.1.1. Narrowband LOS Algorithm (Si5316, Si5323, Si5366)
The LOS circuitry divides down each input clock to produce an 8 kHz to 2 MHz signal. (For the Si5316, the output
of divider N3 (See Figure 1) is used.) The LOS circuitry over samples this divided down inpu t clo ck using a 40 MHz
clock to search for extended periods of time without input clock transitions. If the LOS monitor detects twice the
normal number of samples without a clock edge, a LOSn_INT alarm is declared. Table 8, “AC Characteristics—All
Devices” gives the minimum and maximum amount of time for the LOS monitor to trigger.
6.9.1.2. Wideband LOS Algorithm (Si5322, Si5365)
Each input clock is divided down to produce a 78 kHz to 1.2 MHz signal before entering the LOS monitoring
circuitry. The same LOS algorithm as described in the above section is then used.
6.9.2. FOS Alarms (Si5365 and Si5366)
If FOS alarms are enabled (See Table 32), the internal frequency offset alarms (FOSn_INT) indicate if the input
clocks are within a specified frequency band relative to the frequency of CKIN2. The frequency offset monitoring
circuitry compares the frequency of th e inp ut clo ck (s) wit h CKIN2. If the frequency offset of an input clock exceeds
a preset frequency offset threshold, an FOS alarm (FOSn_INT) is declared for that clock input. Note that FOS
monitoring is not available on CKIN3 and CKIN4 if CK_CONF = 1. The device supports FOS hysteresis per GR1244-CORE, making the device less susceptible to FOS alarm chattering. A TCXO or OCXO reference clock must
be used in conjunction with either the SMC or Stratum 3/3E settings. Note that wander can cause false FOS
alarms.
Rev. 0.573
Si53xx-RM
Table 32. Frequency Offset Control (FOS_CTL)
FOS_CNTLMeaning
LFOS Disabled.
MStratum 3/3E FOS Threshold (12 ppm)
HSONET Minimum Clock Threshold (48 ppm)
6.9.3. FSYNC Align Alarm (Si5366 and CK_CONF = 1 and FRQTBL = L)
At power-up or any time after the PLL has lost lock and relo cked, th e device a utomatica lly perform s a realign ment
of FS_OUT using the currently active sync input. After this, as long as the PLL remains in lock and a realignment is
not requested, FS_OUT will include a fixed number of high-speed clock cycles, even if input clock switches are
performed. If many clock switches are performed, it is possible that the input sync to output sync phase relationship
will shift due to the accumulated residual phase transients of the phase build-out circuitry. The internal ALIGN_INT
signal is asserted when the accumulated phase errors exceeds two cycles of CKOUT2.
6.9.4. C1B and C2B Alarm Outputs (Si5316, Si5322, Si5323)
The alarm outputs (C1B and C2B) are determined directly by the LOS1_INT and LOS2_INT internal indicators
directly. That is C1B = LOS1 and C2B = LOS2.
6.9.5. C1B, C2B, C3B, and ALRMOUT Outputs (Si5365, Si5366)
The alarm outputs (C1B, C2B, C3B, ALRMOUT) provide a summary of various ala rm conditions on the input clocks
depending on the setting of the FOS_CNTL and CK_CONF pins.
The following internal alarm indicators are used in determining the output alarms:
LOSn_INT: See section “6.9.1. Loss-of-Signal Alarms (Si5316, Si5322, Si5323, Si5365, Si5366)” for a
description of how LOSn_INT is determined
FOSn_INT: See section “6.9.2. FOS Alarms (Si5365 and Si5366)”for a description of how FOSn_INT is
determined
ALIGN_INT: See section “6.9.3. FSYNC Align Alarm (Si5366 and CK_CONF = 1 and FRQTBL = L)” for a
description of how ALIGN_INT is determined
Based on the above internal signals and the settings of the CK_CONF and FOS_CTL pins, the outputs C1B, C2B,
C3B, ALRMOUT are determined (SeeTable 33). For details, see "Appendix D—Alarm Structure" on page 144.
.
Table 33. Alarm Output Logic Equations
CK_CONFFOS_CTLAlarm Output Equations
0
Four independent input
clocks
1
(FSYNC switching
mode)
74Rev. 0.5
(Disables FOS)
(Disables FOS)
L
M or HC1B = LOS1_INT or FOS1_INT
C2B = LOS2_INT or FOS2_INT
C3B = LOS3_INT or FOS3_INT
ALRMOUT = LOS4_INT or FOS4_INT
L
M or HC1B = LOS1_INT or (LOS3_INT and FSYNC_SWTCH) or FOS1_INT
C2B = LOS2_INT or (LOS4_INT and FSYNC_SWTCH) or FOS2_INT
C1B = LOS1_INT or (LOS3_INT and FSYNC_SWTCH)
C2B = LOS2_INT or (LOS4_INT and FSYNC_SWTCH)
C1B = LOS1_INT
C2B = LOS2_INT
C3B = LOS3_INT
ALRMOUT = LOS4_INT
C3B = tri-state
ALRMOUT = ALIGN_INT
C3B = tri-state
ALRMOUT = ALIGN_INT
Si53xx-RM
6.9.5.1. PLL Lock Detect (Si5316, Si5323, Si5366)
The PLL lock detection algorithm indicates the lock status on the LOL output pin. The algorithm works by
continuously monitoring the phase of the input clock in relation to the phase of the feedback clock. If the time
between two consecutive phase cy cle slips is greater than the Retrigger Time, the PLL is in lock. The LOL output
has a guaranteed minimum pulse width as shown in (Table 8, “AC Characteristics—All Devices”). The LOL pin is
also held in the active state during an internal PLL calibration.
The retrigger time is automatically set based on the PLL closed loop bandwidth (SeeTable 34).
A PLL loss of lock indicator is not available for these devices.
6.10. Device Reset
Upon powerup, the device internally executes a power-on-reset (POR) which resets the internal device logic. The
pin RST
then performs a PLL Self-Calibration (See “6.2. PLL Self-Calibration”).
can also be used to initiate a reset. The device stays in this state until a valid CKINn is present, when it
6.11. DSPLLsim Configuration Software
To simplify frequency planning, loop bandwidth selection, and general device configuration of the Any-Frequency
Precision Clocks. Silicon Laboratories offers the DSPLLsim configuration utility for this purpose. This software is
available to download from www.silabs.com/timing.
The devices in this family provide a rich set of clock multiplication/clock division options, loop bandwidth selections,
output clock phase adjustment, and device control options.
7.1. Clock Multiplication
The input frequency, clock multiplication ratio, and output frequency are set via register settings. Because the
DSPLL dividers settings are directly programmable, a wide range of fr equency translations is available. In addition,
a wider range of frequency translations is available in narrowband parts than wideband parts due to the lower
phase detector frequency range in narrowband p a rts. To assist users in fin ding valid divider settin gs for a particular
input frequency and clock multiplication ratio, Silicon Laboratories offers the DSPLLsim utility to calculate these
settings automatically. When multiple divider combinations produce the same output frequency, the software
recommends the divider settings that yield the best combination of phase noise performance and power
consumption.
These devices operate as wideband clock multipliers without an external resonator or reference clock. This mode
may be desirable if the input clock is already low jitter and only simple clock multiplication is required. A limited
selection of clock multiplication factors is available in this mode. The input-to-output skew for wideband p arts is not
controlled.
Refer to Figure 25. The selected input clock passes through the N3 input divider and is provided to the DSPLL. The
input-to-output clock multiplication ratio is defined as follows:
Because there is only one DCO and all of the outputs must be frequencies that are integer divisions of the DCO
frequency, there are restrictions on the ratio of one output frequency to another output frequency. That is, there is
considerable freedom in the ratio between the input frequency and the first output frequency; but once the first
output frequency is chosen, there are rest rictions on subsequent output frequencies. These restrictions are made
tighter by the fact that the N1_HS divider is shared among all of the outputs. DSPLLsim should be used to
determine if two different simultaneous outputs are compatible with one another.
The same issue exists for inputs of different freque ncie s: both inputs, after having been divided by their respective
N3 dividers, must result in the same f3 frequency because the phase/frequency detector can operate at only one
frequency at one time.
7.1.2.1. Loop Bandwidth (Si5325, Si5367)
The loop bandwidth (BW) is digit ally progr ammable using the BWSEL_REG[3:0] register bits. The device operating
frequency should be determined prior to loop bandwidth configuration because the loop bandwidth is a function of
the phase detector input frequency and the PLL feedback divider. See DSPLLsim for BWSEL_REG settings and
associated bandwidth.
7.1.2.2. Lock Detect (Si5325, Si5367)
A PLL loss of lock indicator is not available in these devices.
7.1.2.3. Input to Output Skew (Si5325, Si5367)
The input to output skew for wideband devices is not controlled.
The DCO uses the reference clock on the XA/XB pins (OSC_P and OSC_N for the Si5374 and Si5375) as its
reference for jitter attenuation. The XA/XB pins support either a crystal oscillator or an input buffer (single-ended or
differential) so that an external oscillator can become the reference source. In both cases, there are wide margins
in the absolute frequency of the reference input because it is a fixed fre quency and is used only as a jitter reference
and holdover reference (see "7.6. Digi tal Hold" on page 87). See " Appendix A—Narr owb and Refer ences" on page
119 for more details. The Si5374 and Si5375 must be used with an external crystal oscillator and cannot use
crystals.
Care must be exercised in certain areas for optimum performance. For details on this subject, refer to "Appendix
B—Frequency Plans and Jitter Performance (Si5316, Si5319, Si5323, Si5324, Si5326, Si5327, Si5366, Si5368,
Si5369, Si5374, Si5375)" on page 121. For examples of connections to the XA/XB (for the Si5374 and Si5375,
OSC_P, OSC_N) pins, refer to "8.4. Crystal/Reference Clock Interfaces (Si5316, Si5319, Si5323, Si5324, Si5326,
Si5327, Si5366, Si5368, Si5369, Si5374, and Si5375)" on page 113.
Refer to Figure 26 Narrowband PLL Divider Settings (Si5319, Si5324, Si5326, Si5327, Si5368, Si5374, Si5375), a
simplified block diagram of the device and Table 35 and Table 36 for frequency and divider limits. The PLL dividers
and their associated rang es a re lis te d in the diagram. Each PLL divider setting is programmed by writing to device
registers. There are additional restrictions on the range of the input frequency f
rate f3, and the DSPLL output clock f
OSC
.
The selected input clock passes through the N3 input divider and is provided to the DSPLL. In addition, the
external crystal or reference clock provides a reference frequency to the DSPLL. The DSPLL output frequency,
, is divided down by each output divider to generate the clock output frequencies. The input-to-output clock
The output divider, NC1, is the product of a high-speed divider (N1_HS) and a low-speed divider (N1_LS).
Similarly, the feedback divider N2 is the product of a high-speed divider N2_HS and a low-speed divider N2_LS.
When multiple combinations of high-speed and low-speed divider values are available to produce the desired
overall result, selecting the largest possible high-speed divider value will produce lower power consumption. With
the f
ranges from (4 x 220) to 6. For NC1 = 5, the output frequency range 970 MHz to 1.134 GHz can be obtained. For
NC1 = 4, the output frequency range from 1.2125 to 1.4175 GHz is available.
Because there is only one DCO and all of the outputs must be frequencies that are integer divisions of the DCO
frequency, there are restrictions on the ratio of one output frequency to another output frequency. That is, there is
considerable freedom in the ratio between the input frequency and the first output frequency; but once the first
output frequency is chosen, there are restrictions on subsequen t output freq uencies. These restriction s are caused
by the fact that the N1_HS divider is shared among all of the outputs. DSPLLsim should be used to determine if two
different simultaneous outputs are compatible with one another.
The same issue exists for inputs of different frequency: both inputs, after having been divided by their respective
N3 dividers, must result in the same f3 frequency because the phase/frequency detector can operate at only one
frequency at one time.
The device functions as a jitter attenuator with digitally programmable loop bandwidth (BW). The loop bandwidth
settings range from 60 Hz to 8.4 k Hz and are set using the BWSEL_REG[3:0] register bits. The device operating
frequency should be determined prior to loop bandwidth configuration because the loop bandwidth is a function of
the phase detector input frequency and the PLL feedback divider. See DSPLLsim for a table of BWSEL_REG and
associated loop bandwidth settings. For more information the loop BW and its effect on jitter attenuation, see
"Appendix H—Jitter Attenuation and Loop BW" on page 164.
The loop BW of the Si5324, Si5327, Si5369, and Si5374 is significantly lower than the BW of the Si5326. The
available Si5324/27/69/74 loop bandwidth settings and their register control values for a given frequency plan are
listed by DSPLLsim (Revision 4.0.1 or higher) or in Si537xDSPLLsim. Compared to the Si5326, the BW Si5324/2 7/
69/74 settings are approximately 16 times lower, which means that the Si5324/27/69/74 loop bandwidth ranges
from about 4 to 525 Hz.
The device has a PLL lock detection algorithm that indicates the lock status on the LOL output pin and the
LOL_INT read-only register bit. See Section“7.11.8. LOL (Si5319, Si5324, Si5326, Si5327, Si5368, Si5369,
Si5374, Si5375)” for a detailed description of the LOL algo rithm .
and N1 ranges given above, any output frequency can be achieved from 2 kHz to 945 MHz where NC1
OSC
7.2. PLL Self-Calibration
The device performs an internal self-calibration before operation to optimize loop parameters and jitter
performance. While the self-calibration is being performed, the DCO is being internally controlled by the selfcalibration state machine, and the LOL alarm will be active. The output clocks can either be active or disabled
depending on the SQ_ICAL bit setting. The self-calibration time t
All Devices”. The procedure for initiating the internal self-calibration is described below.
7.2.1. Initiating Internal Self-Calibration
Any of the following events will trigger an automatic self-calibration:
Internal DCO registers out-of-range, indicating the need to relock the DCO
Setting the ICAL register bit to 1
In any of the above cases, an internal self-calibration will be initiated if a valid input clock exists (no input alarm)
and is selected as the active clock at that time. The external crystal or reference clock must also be present for the
self-calibration to begin (LOSX_INT = 0 [narrowband only]).
When self-calibration is initiated the device generates an output clock if the SQ_ICAL bit is set to 0. The output
clock will appear when the device begins self-calibration. The frequency of the output clocks may be as high as 5%
above or as low as 20% below the final locked value. If SQ_ICAL = 1, the output clocks are disabled during self-
Rev. 0.579
LOCKMP
is given in Table 8, “AC Characteristics—
Si53xx-RM
calibration and will appear after the self-calibration routine is completed. The SQ_ICAL bit is self-clearing after a
successful ICAL.
After a successful self-calibration has been perform ed with a valid input clock, it is not necessary to reinitiate a selfcalibration for subsequent losses of input clock. If the input clock is lost following self-calibration, the device enters
digital hold mode. When the input clock returns, the device relocks to the input clock without performing a selfcalibration.
After power-up and writing of dividers or PLL registers, the user must set ICAL = 1 to initiate a self-calibration. LOL
will go low when self calibration is complete. Depending on the selected value of the loop bandwidth, it may take a
few seconds more for the output frequency and phase to completely settle.
It is recommended that a software reset precede all ICALs and their associated register writes by setting
RST_REG (Register 136.7).
Due to the low loop bandwidth of the Si5324, Si5327, Si5369, and Si5374, the lock time of the Si5324/27/69/75 is
significantly longer than the lock time of the Si5326. As a method of reducing the lock time, the FAST_LOCK
register bit can be set to improve lock times. As the Si5324/27/69/74 data sheets indicate, FAST_LOCK is the LSB
of register 137. When FAST_LOCK is high, the lock time decreases. Because the Si5324/27/74 is initialized withFAST_LOCK low, it must be written before ICAL is set. Typical Si5324/27/69/74 lock times (as defined from the
start of ICAL until LOL going low) with FAST_LOCK set are one to two seconds. To further reduce lock times, it is
also recommended that a value of 001 be written to LOCKT (the three LSBs of register 19).
7.2.2. Input Clock Stability during Internal Self-Calibration
An ICAL must occur when the selected active CKINn clock is stable in frequency and with a frequ ency value that is
within the operating range that is reported by DSPLLsim. The other CKINs must be stable in frequency (< 100 ppm
from nominal) or squelched during an ICAL.
7.2.3. Self-Calibration Caused by Changes in Input Frequency
If the selected CKINn varies by 500 ppm or more in frequency since the last calibration, the device may initiate a
self-calibration.
The input-to-output skew is not controlled. External circuitry is required to control the input-to-o utput skew. Contact
Silicon Labs for further information.
7.2.5. Clock Output Behavior Before and During ICAL
Table 37. CKOUT_ALWAYS_ON and SQ_ICAL Truth Table
CasesCKOUT_ALWAYS_ONSQ_ICALResults
1
1
2
2
3
3
4
4
Notes:
1. Case 1 should be selected when an output clock is not desired until the part has been initialized after power-up, but is
desired all of the time after initialization.
2. Case 2 should be selected when an output clock is never desired during an any ICAL . Case 2 will only generate
outputs when the outputs are at the correct output frequency.
3. Case 3 should be selected when ever a clock output is always desired.
4. Case 4 is the same as Case 3.
80Rev. 0.5
00CKOUT OFF until after the first ICAL
01
CKOUT OFF until after the first successful
ICAL (i.e., when LOL is low)
10CKOUT always ON, including during an ICAL
CKOUT always ON, including during an ICAL.
11
Use these settings to preserve output-to-output
skew
Si53xx-RM
CKIN1
CKIN2
Clock priority logic
CK_PRIORn
0
1
CKSEL_REG
AUTOSEL_REG0
1
CKSEL_PIN
LOS/FOS
detect
LOS/FOS
detect
LOS/FOS
detect
LOS/FOS
detect
decode
Auto
Manual
Selected
Clock
2
4
CS_CA pin
CK_ACTV_PIN
7.3. Input Clock Configurations (Si5367 and Si5368)
The device supports two input clock configurations based on CK_CONFIG_REG. See "6.5. Frame Synchronization
(Si5366)" on page 70 for additional deta ils.
7.4. Input Clock Control
This section describes the clock selection capabilities (manual input selection, automatic input selection, hitless
switching, and revertive switching). The Si5319, Si5327, and Si5375 support only pin-controlled manual clock
selection. Figure 27 and Figure 28 provide top level overviews of the clock selection logic, though they do not
cover wideband or frame sync applic ations. Register valu es are indicated by underscored italics. Note that, when
switching between two clocks, LOL may temporarily go high if the clocks dif fe r in frequen cy by mo re than 1 00 ppm.
Manual control of input clock selection is available by setting the AUTOSEL_REG[1:0] re gister bit s to 0 0. In manual
mode, the active input clock is chosen via the CKSEL_REG[1:0] register setting according to Table 38 and
Table 39.
CKSEL_REG[1:0]
Register Bits
Note: Setting the CKSEL_PIN register bit to one allows the CS [1:0] pins to continue to control input clock selection.
If the selected clock enters an alarm condition, the PLL enters digital hold mode. The CKSEL_REG[1:0] controls
are ignored if automatic clock selection is enabled.
Automatic switching is either revertive or non-revertive. The default prioritization of clock inputs when the device is
configured for automatic switching operation is CKIN1, followed by CKIN2, and finally, digital hold mode. The
inverse input clock priority arrangement is available through the CK_PRIOR bits, as shown in the Si5325, Si5326,
and Si5374.
For the default priority arrangement, automatic switching mode selects CKIN1 at powerup, reset, or when in
revertive mode with no alarms present on CKIN1. If an alarm condition occurs on CKIN1 and there are no active
alarms on CKIN2, then the device sw itches to CKIN2. If both CKIN1 and CKIN2 are alarmed, then the device
enters digital hold mode. If automatic mode is selected and the frequency offset alarms (FOS1_INT and
FOS2_INT) are disabled, automatic switching is not initiated in response to FOS alarms. The loss-of-signal alarms
(LOS1_INT and LOS2_INT) are always used in making automatic clock selection choices.
In non-revertive mode, once CKIN2 is selected, CKIN2 selection remains as long as it is valid even if alarms are
cleared on CKIN1.
The prioritization of clock inputs for automatic switching is shown in Table 41. For example, if
CK_CONFIG_REG = 0 and the desired clock priority order is CKIN4, CKIN3, CKIN2, and then CKIN1 as the
lowest priority clock, the user should set CK_PRIOR1[1:0] = 11, CK_PRIOR2[1:0] = 10, CK_PRIOR3[1:0] = 01, and
CK_PRIOR4[1:0] = 00.
Table 41. Input Clock Priority for Auto Switching
Selected Clock
CK_PRIORn
If CK_CONFIG_REG = 1 and the desire d clock priority is CKIN1/CKIN3 and then CKIN2/C KIN4, the user should
set CK_PRIOR1[1:0] = 00 and CK_PRIOR2[1:0] = 01 (CK_PRIOR3[1:0] and CK_PRIOR4[1:0] are ignored in this
case).
The following discussion describes the clock selection algorithm for the case of four possible input clocks
(CK_CONFIG_REG = 0) in the default priority arrangement (priority order CKIN1, CKIN2, CKIN3, CKIN4).
Automatic switching mode selects CKIN1 at powerup, reset, or when in revertive mode with no alarms present on
CKIN1. If an alarm condition occurs on CKIN1 and there are no active alarms on CKIN2, the device switches to
CKIN2. If both CKIN1 and CKIN2 are alarmed and there is no alarm on CKIN3, the device switches to CKIN3. If
CKIN1, CKIN2, and CKIN3 are alarmed and there is no alarm on CKIN4, the device switches to CKIN4. If alarms
exist on CKIN1, CKIN2, CKIN3, and CKIN4, the device enters digital hold mode. If automatic mode is selected and
the frequency offset alarms (FOS1_INT, FOS2_INT, FOS3_INT, FOS4_INT) are disabled, au tomatic switching is
not initiated in response to FOS alarms. The loss-of-signal alarms (LOS1_INT, LOS2_INT, LOS3_INT, LOS4_INT)
are always used in making automatic clock selection choices. In non-revertive mode, once CKIN2 is selected,
CKIN2 selection remains as long as it is valid even if alarms are cleared on CKIN1.
Silicon Laboratories switching technology performs phase build-out, which maintains the phase of the output when
the input clock is switched. This minimi zes the propagation of phase transients to the clock outputs during input
clock switching. All switching between input clocks occurs within the input multiplexer and phase detector circuitry.
The phase detector circuitry continually monitors the phase difference between each input clock and the DSPLL
output clock, f
so that the phase offset is maintained by the PLL circuitry.
At the time a clock switch occurs, the phase detector circu itr y kn ows bo th th e inpu t- to -o utpu t p ha se re lation sh ip for
the original input clock and for the new input clock. The phase detector circuitry locks to the new input clock at the
new clock's phase offset so that the phase of the output clock is not disturbed. The phase difference between the
two input clocks is absorbed in the phase detector's offset value, rather than being propagated to the clock output.
The switching technology virtually eliminates the output clock phase transients traditionally associated with clock
rearrangement (input clock switching).
Note that hitless switching between input clocks applies only when the input clock validation time is
VALTIME[1:0] = 01 or higher.
. The phase detector circuitry can lock to a clock signal at a specified phase of fset rela tive to f
OSC
[1:0]CK_CONFIG_REG = 0CK_CONFIG_REG = 1
00CKIN1CKIN1/CKIN3
01CKIN2CKIN2/CKIN4
10CKIN3Not Used
11CKIN4Not Used
7.5. Si5319, Si5324, Si5326, Si5327, Si5368, Si5369, Si5374 and Si5375 Free Run Mode
Figure 29. Free Run Mode Block Diagram
CKIN2 has an extra mux with a path to the crystal oscillator output.
When in Free Run mode, CKIN2 is sacrificed (Si5326, Si5368, Si5369, Si5374).
Switching between the crystal oscillator and CLKIN1 is hitless.
Either a crystal or an external oscillator can be used.
External oscillator connection can be either single ended or differential.
All other features and specifications remain the same.
7.5.1. Free Run Mode Programming Procedure
Using DSPLLsim, determine the frequency plan:
Write to the internal dividers, including N31 and N32.
Enable Free Run Mode (the mux select line), FREE_RUN.
Select CKIN1 as the higher priority clock.
Establish revertive an d au to se lec t mo de s.
Once properly programmed, the part will:
Initially lock to either the XA/XB (OSC_P and OSC_N for the Si5374/75) or to CKIN1.
Automatically select CKIN1, if it is available.
Automatically and hitlessly switch to XA/XB if CKI N1 fails.
Automatically and hitlessly switch back to CKIN1 when it subsequentl y returns.
For the Si5319:
Clock selection is manual using an input pin.
Clock switching is not hitless.
CKIN2 is not available.
7.5.2. Clock Control Logic in Free Run Mode
Noting that the mux that selects CKIN2 versus the XA/XB oscillator is located before the clock selection and control
logic, when in Free Run mode operation, all such logic will be driven by the XA/XB oscillator, not the CKIN2 pins.
For example, when in Free Run mode, the CK2B pin will reflect the status of the XA/XB oscillator and not the status
of the CKIN2 pins.
Rev. 0.585
Si53xx-RM
CKIN
N31
-------------- -
XA-XB
N32
----------------- -
f
3
==
CKOUT
XA-XB
--------------------- -
Integer
7.5.3. Free Run Reference Frequency Constraints
XA/XB Frequency MinXA/XB Frequency MaxXtal
109 MHz125.5 MHz3rd overtone
37 MHz41 MHzFundamental
All crystals and external oscillators must lie within these two bands
Not every crystal will work; they should be tested
An external oscillator can be used at all four bands
The frequency at the phase detector (f3) must be the same for both CKIN1 and XA/XB or else switching cannot
be hitless
To avoid spurs, avoid outputs that are an integer (or near integer) of the XA/XB frequency.
7.5.4. Free Run Reference Frequency Constraints
While in Free Run:
CKOUT frequency tracks the reference frequency.
For very low drift, a TCXO or OCXO reference is necessary.
CKOUT Jitter:
XA/XB to CKOUT jitter transfer function is roughly one-to-one.
For very low jitter, either use a high quality crystal or external oscillator.
3rd overtone crystals have lower close-in phase noise.
In general, higher XA/XB frequency > lower jitter.
XA/XB frequency accuracy:
For hitless switching, to meet all published specifications, the XA/XB frequency divided by N32 should match the CLKIN
frequency divided by N31. If they do not match, the clock switch will still be well-behaved.
Other than the above, the absolute accuracy of the XA/XB frequency is not important.
86Rev. 0.5
Si53xx-RM
Time
Digital Hold
@
t = 0
MM
HIST
t = –HIST_DEL
HIST_AVG
7.6. Digital Hold
All Any-Frequency Precision Clock devices feature a holdover mode, whereby the DSPLL is locked to a digital
value.
7.6.1. Narrowband Digital Hold (Si5316, Si5324, Si5326, Si5368, Si5369, Si5374)
After the part's initial self-calibration (ICAL), when no valid input clock is available, the device enters digital hold.
Referring to the logical diagram in "Appendix D—Alarm Structure" on page 144, lack of clock availability is defined
by following the boolean equation for the Si5324, Si5326, and Si5374:
(LOS1_INT OR FOS1_INT) AND (LOS2_INT OR FOS2_INT) = enter digital hold
The equivalent Boolean equation for the Si5327 is as follows:
LOS1 and LOS2 = enter digital hold
The equivalent boolean equation for the Si5367, Si5368, and Si5369 is as follows:
(LOS1_INT OR FOS1_INT) AND (LOS2_INT OR FOS2_INT) AND
(LOS3_INT OR FOS3_INT) AND (LOS4_INT OR FOS4_INT) = enter digital hold
7.6.1.1. Digital Hold Detailed Description (Si5324, Si5326, Si5327, Si5368, Si5369, Si5374)
In this mode, the device provides a stable output fr equency until the input clock returns and is validated. Upon
entering digital hold, the internal DCO is initially held to its last frequency value, M (See Figure 30). Next, the DCO
slowly transitions to a historical average frequency value supplied to the DSPLL, M
Values of M starting from time t = –(HIST_DEL + HIST_AVG) and ending at t = –HIST_DEL are averaged to
compute M
record of previous M values supplied to the DCO. By using a historical average frequency, input clock phase and
frequency transients that may o ccur immediately preceding digital hold do not affect the digital hold frequency.
Also, noise related to input clock jitter or internal PLL jitter is minimized.
. This historical average frequency value is taken from an internal memory location that keeps a
HIST
, as shown in Figure 30.
HIST
Figure 30. Parameters in History Value of M
The history delay can be set via the HIST_DEL[4:0] register bits as shown in Table 42 and the history averaging
time can be set via the HIST_AVG[4:0] register bits as shown in Table 43. The DIGHOLDVALID register can be
used to determine if the information in HIST_AVG is valid and the device can enter SONET/SDH compliant digital
hold. If DIGHOLDVALID is not active, the part will enter VCO freeze instead of digital hold.
Rev. 0.587
Si53xx-RM
Table 42. Digital Hold History Delay
HIST_DEL[4:0]History Delay Time (ms)HIST_DEL[4:0]History Delay Time (ms)
If a highly stable reference, such as an oven-controlled crystal oscillator (OCXO) is supplied at XA/XB, an
extremely stable digital hold can be achieved. If a crystal is supplied at the XA/XB port, the digital hold stability will
be limited by the stability of the crystal.
88Rev. 0.5
Si53xx-RM
Normal operation
Input clock drifts
f
0
freq
LOS alarm occurs,
Start Digital hold
Digital Hold
VCO freeze
HIST_AVGHIST_DEL
Clock input
cable is pulled
time
~1 sec
7.6.2. History Settings for Low Bandwidth Devices (Si5324, Si5327, Si5369, Si5374)
Because of the extraordinarily low loop bandwidth of the Si5324, Si5369 and Si5374, it is recommended that the
values for both history registers be increased for longer histories.
7.6.3. Recovery from Digital Hold (Si5319, Si5324, Si5326, Si5327, Si5368, Si5369, Si5374)
When the input clock signal returns, the device transitions from digital hold to the selected input clock. The device
performs hitless recovery from digital hold. The clock transition from digital hold to the returned input clock includ es
“phase buildout” to absorb the phase difference between the digital hold clock phase and the input clock phase.
If an LOS or FOS condition exists on the selected input clock, the device enters VCO freeze. In this mode, the
device provides a stable output frequency until the input clock returns and is validated. When the device enters
digital hold, the internal oscillator is initially held to the frequency value at roughly one second prior to the leading
edge of the alarm condition. VCO freeze is not compliant with SONET/SDH MTIE requirements; applications
requiring SONET/SDH MTIE requirements should use the Si5324, Si5326, Si5368, Si5369 or Si5374. Unlike the
Si5325 and Si5367, the Si5319’s VCO freeze is controlled by the XA/XB reference (which is typically a crystal)
resulting in greater stability. For the Si5319, Si5327, and Si5375, VCO freeze is similar to the Digital Hold function
of the Si5326, Si5368, and Si5369 except that the HIST_AVG and HIST_DEL registers do not exist.
7.6.5. Digital Hold versus VCO Freeze
Figure 31 below is an illustration of the difference in behavior between Digital Hold and VCO Freeze.
Figure 31. Digital Hold vs. VCO Freeze Example
Rev. 0.589
Si53xx-RM
7.7. Output Phase Adjust (Si5326, Si5368)
The device has a highly accurate, digitally controlled device skew capability. For more information on Output Phase
Adjustments, see both DSPLLsim and the respective data sheets. Both can be downloaded by going to
www.silabs.com/timing and clicking on “Documentation” at the bottom of the page.
7.7.1. Coarse Skew Control (Si5326, Si5368)
With the INCDEC_PIN register bit set to 0 (pin control off), overall device skew is controlled via the CLAT[7:0]
register bits. This skew control has a resolution of 1/f
25.4 ns. Following a powerup or reset (RST
Any further changes made in the skew register will be read and compared to the previously held value. The
difference will be calculated and applied to the clock outputs. All skew changes are made in a glitch-free fashion.
When a phase adjustment is in progress, any new CLAT[7:0] values are ignored until the update is complete. TheCLAT PROG register bit is set to 1 during a coarse skew adjustment. The time for an adjustment to complete is
dependent on bandwidth and the delta value in CLAT. To verify a written value into CLAT, the CLAT register should
be read after the register is written. The time that it takes for the effects of a CLAT change to complete is
proportional to the size of the change, at 83 msec for every unit change, assuming the lowest available loop
bandwidth was selected. For example, if CLAT is zero and has the value 100 written to it, the changes will
complete in
100 x 83 msec = 8.3 sec.
If it is necessary to set the high-speed output clock divider N1_HS to divide-by-4 in order to achieve the desired
overall multiplication ratio and output frequency, only phase increments are allowed and negative settings in the
CLAT register or attempts to decrement the phase via writes to the CLAT register will be ignored. Because of this
restriction, when there is a choice between using N1_HS = 4 and another N1_HS value that can produce the
desired multiplication ratio, the other N1_HS value should be selected. This restriction also applies when using the
INC pin.
With the INCDEC_PIN register bit set to 1 (pin control on), the INC and DEC pins function the same as they do for
pin controlled part s. See "6.6. Output Phase Adjust (Si5323, Si5366)" on page 71.
Using the following procedure, the CLAT register can be used to adjust the device clock output phase to an
arbitrarily large value that is not limited by the size of the CLAT register:
1. Write a phase adjustment value to the CLAT register (Register 16). The DSPLLsim configuration software
provides the size of a single step.
2. Wait until CLATPROGRESS = 0 (register 130, bit 7), which indicates that the adjustment is complete (Maximum
time for adjustment: 20 seconds for the Si5326 or Si5368).
3. Set INCDEC_PIN = 1 (Register 21, bit 7).
4. Write 0 to CLAT register (Register 16).
5. Wait until CLATPROGRESS =0.
6. Set INCDEC_PIN =0.
7. Repeat the above process as many times as desired.
Steps 3-6 will clear the CLAT register without changing the output phase. This allows for unlimited output clock
phase adjustment using the CLAT register and repeating steps 1–3 as many times as needed.
Note: The INC and DEC pins must stay low during this process.
7.7.2. Fine Skew Control (Si5326, Si5368)
An additional fine adjustment of the overall device skew can be used in conjunction with the INC and DEC pins or
the CLAT[7:0] register bits to provide finer resolution output phase adjustments. Fine phase adjustment is available
using the FLAT[14:0] bits. The nominal range and resolution of the FLAT[1 4:0] skew adjustment word are:
Range FLAT = ±110 ps
Resolution FLAT = 9 ps
pin or RST_REG register bit), the skew will revert to the reset value.
, approximately 200 ps, and a range from –25.6 to
OSC
90Rev. 0.5
Si53xx-RM
Before writing a new FLAT[14:0] value, the FLAT_VALID bit must be set to 0 to hold the existing FLAT[14:0] value
while the new value is being written. Once the new value is written, set FLAT_VALID = 1 to enable its use.
To verify a written value into FLAT, the FLAT register should be read after the register is written.
Because the FLAT resolution varies with the frequency plan and selected bandwidth, DSPLLsim reports the FLAT
resolution each time it creates a new frequency plan.
Because of its very low loop bandwidth, the output phase of the Si5324, Si5327, Si5369, and Si5374 are not
adjustable. This means that the Si5324, Si5327, Si5369, and Si5374 do not have any INC or DEC pins and that
they do not have CLAT or FLAT registers.
The phase of each clock output may be adjusted in relation to the phase of the other clock outputs, respectively.
This feature is available when CK_CONFIG_REG = 0. The resolution of the phase adjustment is equal to [NI HS/
]. Since F
F
VCO
800 ps to 2.2 ns depending on the PLL divider settings. Silicon Laboratories' PC-based configuration software
(DSPLLsim) provides PLL divider settings for each frequency translation, if applicable. If more than one set of PLL
divider settings is available, selecting the combination with the lowest N1_HS value provides the finest resolution
for output clock phase offset control. The INDEPENDENTSKEWn[7:0] (n = 1 to 5) register bits control the phase of
the device output clocks. By prog ramming a different phase offset for each output clock, output-to-output delays
can easily be set.
The output-to-output skew is guaranteed to be preserved only if the following two register bits are both high:
Register Bit:Location
CKOUT_ALWAYS_ONaddr 0, bit 5
SQICALaddr 3, bit 4In addition, if SFOUT is changed, the output-to-output skew may be distu r be d un til after a successful ICAL.
Note: CKOUT5 phase is random unless it is used for Frame Sync (See section 7.8).
7.7.5. Input-to-Output Skew (All Devices)
The input-to-output skew for these devices is not controlled.
is approximately 5 GHz and N1_HS = (4, 5, 6, …, 11), the resolution varies from approximately
VCO
7.8. Frame Synchronization Realignment (Si5368 and CK_CONFIG_REG =1)
Frame Synchronization Realignment is selected by setting CK_CONFIG_REG = 1. In a typical frame
synchronization application, CKIN1 and CKIN2 are high-speed input clocks from primary and secondary clock
generation cards and CKIN3 and CKIN4 are their associated primary and secondary frame synchronization
signals. The device generates four output clocks and a frame sync output FS_OUT. CKIN3 and CKIN4 control the
phase of FS_OUT. When CK_CONFIG_REG = 1, the Si5368 can lock onto only CKIN1 or CKIN2. CKIN3 and
CKIN4 are used only for purposes of frame synchronization.
The inputs supplied to CKIN3 and CKIN4 can range from 2 to 512 kHz. So that two different frame sync input
frequencies can be accommodated, CKIN3 and CKIN4 each have their own input dividers, as shown in Figure 32.
The CKIN3 and CKIN4 frequencies are set by the CKIN3RATE[2:0] and CKIN4RATE[2:0] register bits, as shown in
Table 44. The frequency of FS_OUT can range from 2 kHz to 710 MHz and is set using the NC5_LS divider
setting. FS_OUT must divide evenly into CKOUT2. For example, if CKOUT2 is 156.25MHz, then 8 kHz woul d no t
be an acceptable frame rate because 156.25 MHz/8 kHz = 19,531.25, which is not an integer. However, 2 kHz
would be an acceptable frame rate because 156.25 MHz/2 kHz = 78,125.
Rev. 0.591
Si53xx-RM
CKIN3
CKIN4
CLKIN3RATE
Clock select
DCO,
4.85 GHz
to 5.67 GHz
CKOUT2
N1_HS
NC2_LS
NC5_LS
FS_OUT
CLKIN4RATE
to align
Typically the same frequency
CKIN3
CKIN4
CLKIN3RATE
Clock select
DCO,
4.85 GHz
to 5.67 GHz
N1_HS
NC2_LS
NC5_LS
CLKIN4RATE
to align
Typically the same frequency
Table 44. CKIN3/CKIN4 Frequency Selection (CK_CONF = 1)
The NC5_LS divider uses CKOUT2 as its clock input to derive FS_OUT. The limits for the NC5_LS divider are
NC5_LS = [1, 2, 4, 6, …, 2
f
CKOUT2
< 710 MHz
Note that when in frame synchro niz at ion re a lign m en t m o de, writes to NC5_LS are controlled by FPW_VALID. See
section “7.8.4. FS_OUT Polarity and Pulse Width Control (Si5368)”.
Common NC5_LS divider settings on FS_OUT are shown in Table 45.
7.8.1. FSYNC Realignment (Si5368)
The FSYNC_ALIGN_PIN bit determines if the realignment will be pin-controlled via the FS_ALIGN pin or registercontrolled via the FSYNC_ALIGN_REG register bit. The active CKIN3 or CKIN4 edge to be used is controlled via
the FSYNC_POL register bit.
In either FSYNC alignment control mode, the resolution of the phase realignment is 1 clock cycle of CKOUT2. If
the realignment control is not active, the NC5 divider will continuously divide down its f
guarantees a fixed number of high-frequency clock (CKOUT2) cycles between each FS_OUT cycle.
At power-up, the device automatically performs a realignment of FS_OUT using the currently active sync input.
After this, as long as the PLL remains in lock and a realignment is not requested, FS_OUT will include a fixed
number of high-speed clock cycles, even if input clock switches are performed. If many clock switches are
performed in phase build-out mode, it is possible that the input sync to output sync phase relationship will shift due
to the accumulated residual phase transients of the phase build-out circuitry. The ALIGN_ERR[8:0] status register
reports the deviation of the input-to-output sync phase skew from the desired FSYNC_SKEW[16:0] value in units of
f
CKOUT2
periods. A programmable threshol d to trigger the ALIGN_INT alarm can be set via the ALIGN_THR[2:0]
bits, whose settings are given in Table 46. If the sync alignment error exceeds the threshold in either th e positive or
negative direction, the alarm becomes active. If it is then desired to reestablish the desired input-to-output sync
phase relationship, a realignment can be performed. A realignment request may cause FS_OUT to
instantaneously shift its output edge location in order to align with the active input sync phase.
19
]
Table 45. Common NC5 Divider Settings
CKOUT2 Frequency (MHz)NC5 Divider Setting
2 kHz FS_OUT8 kHz FS_OUT
19.4497202430
77.76388809720
155.527776019440
622.0831104077760
CKOUT2
input. This
Table 46. Alignment Alarm Trigger Threshold
ALIGN_THR [2:0]Alarm Trigger Threshold (Units of T
0004
0018
01016
01132
10048
10164
11096
111128
Rev. 0.593
CKOUT2
)
Si53xx-RM
For cases where phase skew is required, see Section “7.7. Output Phase Adjust (Si5326, Si5368)” for mor e det ails
on controlling the sync input to sync output phase skew via the FSYNC_SKEW[16:0] bits. See Section“8.2. Output
Clock Drivers” for information on the FS_OUT signal format, pulse width, and active logic level control.
7.8.2. FSYNC Skew Control (Si5368)
When CKIN3 and CKIN4 are configured as frame sync inputs (CK_CONFIG_REG = 1), phase skew of the sync
input active edge to FS_OUT active edge is controllable via the FSYNC_SKEW[16:0] register bits. Skew control
has a resolution of 1/f
CKOUT2
and a range of 131,071/f
CKOUT2
period of CKIN3, CKIN4, and FS_OUT.
The skew should not be changed more than once per FS_OUT period. If a FSYNC realignment is being m ade, the
skew should not be changed until the realignment is complete. The skew value and the FS_OUT pulse width
should not be changed within the same FS_OU T pe rio d .
Before writing the three bytes needed to specify a new FSYNC_SKEW[16:0] value, the user shou ld set the reg ister
bit FSKEW_VALID = 0. This causes the alignment state machine to keep using the previous FSYNC_SKEW[16:0]
value, ignoring the new register values as they are being written. Once the new FSYNC_SKEW[16:0] value has
been completely written, the user should set FSKEW_VALID = 1 at which time the alignment state machine will
read the new skew alignment value. Note that when the new FSYNC_SKEW[16:0] value is used, a phase step will
occur in FS_OUT.
7.8.3. Including FSYNC Inputs in Clock Selection (Si5368)
The frame sync inputs, CKIN3 and CKIN4, are both monitored for loss-of-signal (LOS3_INT and LOS4_INT)
conditions. To include these LOS alarms in the input clock selection algorithm, set FSYNC_SWTCH_REG =1. The
LOS3_INT is logically ORed with LOS1_INT and LOS4_INT is ORed with LOS2_INT as inputs to the clock
selection state machine. If it is desired not to include these alarms in the clock selection algorithm, set
FSYNC_SWTCH_REG = 0. The frequency offset (FOS) alarms for CKIN1 and CKIN2 can also be included in the
state machine decision making as described in Section “7.11. Alarms (Si5319, Si5324, Si5325, Si5326, Si5327,
Si5367, Si5368, Si5369, Si5374, Si5375)”; however, in frame sync mode (CK_CONFIG_REG = 1), the FOS alarms
for CKIN3 and CKIN4 are ignored.
7.8.4. FS_OUT Polarity and Pulse Width Control (Si5368)
Additional output controls are available for FS_OUT. The active polarity of FS_OUT is set via the FS_OUT_POL
register bit and the active duty cycle is set via the FSYNC_PW[9:0] register. Pulse width settings have a resolution
of 1/f
CKOUT2
, and a 50% duty cycle setting is provided. Pulse width settings can range from 1 to (NC5-1) CKOUT2
periods, providing the full range of pulse width possibilities for a given NC5 divider setting.
The FS_OUT pulse should not be changed more than once per FS_OUT period. If a FSYNC realignment is being
made, the pulse width should not be changed until the realignment is complete. The FS_OUT pulse width and the
skew value should not be changed within the same FS_OUT period.
Before writing a new value into FSYNC_PW[9:0], the user should set the register bit FPW_VALID = 0. This causes
the FS_OUT pulse width state machine to keep using the previous FSYNC_PW[9:0] value, ignoring the new
register values as they are being written. Once the new FSYNC_PW[9:0] value has been completely written, the
user should set FPW_VALID = 1, at which time the FS_OUT pulse width state machine will read the new pulse
width value.
Writes t o NC5_LS should be treated the same as writes to FSYNC_PW. Thus, all writes to NC5_LS should occur
only when FPW_VALID = 0. Any such writes will not take effect until FPW_VALID = 1.
Note that f
CKOUT2
must be less than or equal to 710 MHz when CK_CONFIG_REG = 1; otherwise, the FS_OUT
buffer and NC5 divider must be disabled.
7.8.5. Using FS_OUT as a Fifth Output Clock (Si5368)
In applications where the frame synchronization functionality is not needed (CK_CONFIG_REG = 0), FS_OUT can
be used as a fifth clock output. In this case, no realignment requests should be made to the NC5 divider (hold
FS_ALIGN = 0 and FSYNC_ALIGN_REG = 0). Output pulse width and polarity controls for FS_OUT are still
available as described above. The 50% duty cycle setting would be used to generate a typical balanced output
clock.
The device includes a flexible output driver structure that can drive a variety of loads, including LVPECL, LVDS,
CML, and CMOS formats. The signal format of each output is individually configurable through the
SFOUTn_REG[2:0] register bits, which modify the output common mode and differential signal swing.
Table 47 shows the signal formats based on the supply voltage and the type of load being driven. For the CMOS
setting, both output pins drive single-ended in-phase signals and should be externally shorted together to obtain
the maximum drive strength.
The SFOUTn_REG[2:0] register bits can also be used to disable the outputs. Disabling the outputs puts the
CKOUT+ and CKOUT– pins in a high-impedance state relative to V
outputs remain connected to each other th rou gh a 200 on-chip resistance (differential impedance of 200 ). The
clock output buffers and DSPLL output dividers NCn are powered down in disable mode.
The additional functions of “Hold Logic 1” and “Hold Logic 0”, which create static logic levels at the outputs, are
available. For differential output buffer formats, the H old Logic 1 state causes the positive output of the differential
signal to remain at its high logic level while the negative output remains at the low logic level. For CMOS output
buffer format, both outputs remain high during the Hold Logic 1 state. These functions are controlled by the
HLOG_n bits. When entering or exiting the “Hold Logic 1” or “Hold Logic 0” states, no glitches or runt pulses are
generated on the outputs. Changes to SFOUT or HLOG will change the output phase. An ICAL is required to reestablish the output phase. When SFOUT = 010 for CMOS, bypass mode is not supported.
7.9.1. Disabling CKOUTn
Disabling CKOUTn output powers down the output buffer and output divider. Individual disable controls are
available for each output using the DSBLn_REG.
7.9.2. LVPECL TQFP Output Signal Format Restrictions at 3.3 V (Si5367, Si5368, Si5369)
The LVPECL and CMOS output formats draw more current than either LVDS or CML; therefore, there are
restrictions in the allowed output format pin settings that limit the maximum power dissipa tion for the TQFP devices
when they are operated at 3.3 V. When Vdd = 3.3 V and there are four enabled LVPECL or CMOS output s, the fif th
output must be disabled. When Vdd = 3.3 V and there are five enabled outputs, there can be no more than three
outputs that are either LVPECL or CMOS. All other configurations are valid, including all with Vdd = 2.5 V.
The device supports a PLL bypass mode in which the selected input clock is fed directly to the ou tput buffers,
bypassing the DSPLL. In PLL bypass mode, the input and output clocks will be at the same frequency. PLL bypass
mode is useful in a laboratory en vironment to m easure syst em performanc e with and without the jitt er attenuatio n
provided by the DSPLL. The BYPASS_REG bit controls enabling/disabling PLL bypass mode.
Before going into bypass mode, it i s recommende d that the p art enter Digit al Hold by setting DHOLD. Internally, the
bypass path is implemented with high-speed diff erential signaling for low jitter. Note that the CMOS output format
does not support bypass mode.
Summary alarms are available to indicate the overall status of the input signals and fr ame a lignment (Si53 68 o nly).
Alarm outputs stay high until all the alarm conditions for that alarm output are cleared. The Register VALTIME
controls how long a valid signal is re-applied before an alarm clears. Table 48 shows the available settings. Note
that only for VALTIME[1:0] = 00, hitless switching is not possible.
Table 48. Loss-of-Signal Validation Times
VALTIME[1:0]Clock Validation Time
002 ms
(hitless switching not available)
01100 ms
10200 ms
1113 s
The device has loss-of-signal circuitry that continuously monitors CKINn for missing pulses. The LOS circuitry
generates an internal LOSn_INT output signal that is processed with other alarms to generate CnB and
ALARMOUT.
An LOS condition on CKIN1 causes the internal LOS1_INT alarm become active. Similarly, an LOS condition on
CKINn causes the LOSn_INT alarm become active. Once a LOSn_INT alarm is asserted on one of the input
clocks, it remains asserted until that input clock is validated over a designated time period. If another error
condition on the same input clock is detected during the validation time then the alarm remains asserted and the
validation time starts over.
There are three options for LOS: LOS, LOS_A, and no LOS, which are selected using the LOSn_EN registers. The
values for the LOSn_EN registers are given in Table 49.
Table 49. Loss-of-Signal Registers
LOSn_EN[1:0]LOS Selection
00Disable all LOS monitoring
01Reserved
10LOS_A enabled
11LOS enabled
96Rev. 0.5
Si53xx-RM
7.11.1.2. Standard LOS (Si5319, Si5324, Si5326, Si5327, Si5368, Si5369, Si5374, Si5375)
To facilitate automatic hitless switching, the LOS trigger time can be significantly reduced by using the default LOS
option (LOSn_EN = 11). The LOS circuitry divides down each input clock to produce a 2 kHz to 2 MHz signal. The
LOS circuitry over samples this divided down input clock using a 40 MHz clock to search for extended periods of
time without input clock transitions. If the LOS monitor detects twice the normal number of samples without a clock
edge, an LOS alarm is declared. The LOSn trigger window is based on the value of the inpu t di vider N3. The valu e
of N3 is reported by DSPLLsim.
The range over which LOS is guaranteed to not produce false positive assertions is 100 ppm. For example, if a
device is locked to an input clock on CKIN1, the frequency of CKIN2 should differ by no more than 100 ppm to
avoid false LOS2 assertions.
The frequency range over which FOS monitoring may occur is from 10 to 710 MHz.
A slower response version of LOS called LOSA is available and should be used under certain conditions. Because
LOSA is slower and less sensitive than LOS, its use should be considered for applications with quasi-periodic
clocks (e.g., gapped clocks with one or more consecutive clock edges removed), when switching between input
clocks with a large difference in frequency and any other application where false positive assertions of LOS may
incorrectly cause the Any-Frequency device to be forced into Digital Hold.
For example, it is recommended that while in Free Run Mode LOSA be used instead of LOS beca use the two clock
inputs will not be the same exact frequency. This will avoid false LOS assertions when the XA/XB frequency differs
from the other clock inputs by more than 100 ppm. See Section 7.11.1.3 for more information on LOSA.
For situations where no form of LOS is desired, LOS can be disabled by writing 00 to LOSn_EN. This mode is
provided to support applications which implement custom LOS algorithms off-chip. If this approach is taken, the
only remaining methods of entering Digital Hold will be FOS or by setting DHOLD (register 3, bit 5).
7.11.1.5. Wideband LOS Algorithm (Si5322, Si5365)
Each input clock is divided down to produce a 78 kHz to 1.2 MHz signal before entering the LOS monitoring
circuitry. The same LOS algorithm as described in the above section is then used. FOS is not available in wideband
devices.
When LOS is enabled, an LOS condition on CKI N1 causes LOS1_INT to become active. Similarly, when LOS is
enabled, an LOS condition on CKIN2 causes LOS2_INT to become active. Once a LOSn_INT alarm is asserted on
one of the input clocks, it remains asserted until the input clock is validated over a designated time period. If
another error condition on the same input clock is detected during the validation time then the alarm remains
asserted and the validation time starts over.
The frequency offset (FOS) alarms indicate if the input clocks are within a specified frequency range relative to the
frequency of a reference clock. The reference clock can be provided by any of the four input clocks (two for Si5324,
Si5325 or Si5326) or the XA/XB input. The default FOS reference is CKIN2. The frequency monitoring circuitry
compares the frequency of the input clock(s) with the FOS reference clock If the frequency offset of an input clock
exceeds a selected frequency offset threshold, an FOS alarm (FOS_INT register bit) is declared for that clock
input. Be aware that large amounts of wander can cause false FOS alarms.
Note: For the Si5368, If CK_CONFIG_REG = 1, only CKIN1 and CKIN2 are monitored; CKIN3 and CKIN4 are used for
FSYNC and are not monitored.
The frequency offset threshold is selectable using the FOS_THR[1:0] bits. Settings are available for compatibility
with SONET Minimum Clock (SCMD) or Stratum 3/3E requirements. See Table 8 on page 40. The device supports
FOS hystereses per GR-1244- CORE, making the device less susceptible to FOS alarm chattering. A reference
clock with suitable accuracy and drift specifications to support the intended application should be used. The FOS
reference clock is set via the FOSREFSEL[2:0] bits as shown in Table 50. More than one input can be monitored
against the FOS reference, i.e., the re can be more than one m onitored clock, but only on e FOS reference. When
the XA/XB input is used as the FOS reference, there is only one reference frequency band that is allowed: from
37 MHz to 41 MHz.
Rev. 0.597
Si53xx-RM
CKIN
FOS_REF
P
Q
FOS
Compare
10 MHz min,
27 MHz max
Table 50. FOS Reference Clock Selection
FOS Reference
FOSREFSEL
all othersReservedReserved
Both the FOS reference and the FOS monitored clock mus t be divided down to the same clock rate and this clock
rate must be between 10 MHz and 27 MHz. As ca n b e see n in F igu re 33 , th e va lu es for P an d Q m ust be se lec ted
so that the FOS comparison occurs at the same frequency. The registers that contain the values for P and Q are
the CKINnRATE[2:0] registers.
The frequency band of each input clock must be specified to use the FOS feature. The CLKNRATE registers
specify the frequency of the device input clocks as shown in Table 51.
When the FOS reference is the XA/XB oscillator (either internal or external), the value of Q in Figure 33 is always
2, for an effective CLKINnRATE of 1, as shown in Table 51.
Table 51. CLKnRATE Registers
CLKnRATEDivisor, P or QMin Frequency, MHzMax Frequency, MHz
011027
122554
2450105
3895215
416190435
532375710
For example, to monitor a 544 MHz clock at CKIN1 with a FOS reference of 34 MHz at CKIN2:
CLK1RATE = 5
CLK2RATE = 1
FOSREFSEL[2:0] = 010
A LOS condition causes the associated LOS1_INT or LOS2_INT read only register bit to be set. A LOS condition
on CKIN_1 will also be reflected onto C1B if CK1_BAD_PIN = 1. Likewise, a LOS condition on CKIN_2 will also be
reflected onto C2B if CK2_BAD_PIN =1.
A FOS condition causes the associated FOS1_INT or FOS2_INT read only register bit to be set. FOS monitoring is
enabled or disabled using the FOS_EN bit. If FOS is enabled (FOS_EN = 1) and CK1_BAD_PIN = 1, a FOS
condition will also be reflected onto its associated output pin, C1B or C2B. If FOS is disabled (FOS_EN =0), the
FOS1_INT and FOS2_INT register bits do not af fect the C1B and C2B alarm outputs, respectively.
Once an LOS or FOS alarm is asserted on one of the input clocks, it is held high until the input clock is validated
over a designated time period. The validation time is programmable via the VALTIME[1:0] register bits as shown in
Table 48 on page 96. If another error condition on the same input clock is detected during the validation time then
the alarm remains asserted and the validation time starts over.
[Si5326]: Note that hitless switching between input clocks applies only when the input clock validation time
VALTIME[1:0] = 01 or higher.
7.11.4. LOS (Si5319, Si5375)
A LOS condition causes the LOS_INT read only register bit to be set. This LOS condition will also be reflected onto
the INT_CB pin.
The generation of alarms on the C1B, C2B, C3B, and ALRMOUT outputs is a function of the input clock
configuration, and the frequency offset alarm enable as shown in Table 52. The LOSn_INT and FOSn_INT signals
are the raw outputs of the alarm monitors. These appear directly in the device status registers. Sticky versions of
these bits (LOSn_FLG, FOSn_FLG) drive the output interrupt and can be individually masked. When the device
inputs are configured as four input clocks (CK_CONFIG = 0), the ALRMOUT pin reflects the status of the CKIN4
input. The equations below assume that the output alarm is active high; however, the active polarity is selectable
via the CK_BAD_POL bit.
Operation of the C1B, C2B, C3B, and ALRMOUT pins is enabled based on setting the C1B_PIN, C2B_PIN,C3B_PIN, and ALRMOUT_PIN register bits. Otherwise, the pin will tri-state. Also, if INT_PIN = 1, the interrupt
functionality will override the appearance of ALRMOUT at the output even if ALRMOUT_PIN =1.
Once an LOS or FOS alarm is asserted for one of the input clocks, it is held high until the input clock is validated
over a designated time period. The validation time is programmable via the VALTIME[1:0] register bits as shown in
Table 48 on page 96. If another error condition on the same input clock is detected during the validation time then
the alarm remains asserted and the validation time starts over.
Note that hitless
VALTIME[1:0] = 01 or higher.
For details, see "Appendix D—Alarm Structure" on page 144.
switching betwee
n input clocks applies only when the input clock validation time
The generation of alarms on the C1B, C2B, C3B, and ALRMOUT outputs is a function of the input clock
configuration, and the frequency offset alarm enable as shown in Table 53. The LOSn_INT and FOSn_INT signals
are the raw outputs of the alarm monitors. These appear directly in the device status registers. Sticky versions of
these bits (LOSn_FLG, FOSn_FLG) drive the output interrupt and can be individually masked. Since, CKIN3 and
CKIN4 are configured as frame sync inputs (CK_CONFIG_REG = 1), ALRMO U T fu nc tio ns as th e a lig nm e nt ala rm
output (ALIGN_INT) as described in Section “7.8. Frame Synchronization Realignment (Si5368 and
CK_CONFIG_REG = 1)”. The equations below assume that the output alarm is active high; however, the active
polarity is selectable via the CK_BAD_POL bit.
Operation of the C1B, C2B, C3B, and ALRMOUT pins is enabled based on setting the C1B_PIN, C2B_PIN,C3B_PIN, and ALRMOUT_PIN register bits. Otherwise, the pin will tri-state. Also, if INT_PIN = 1, the interrupt
functionality will override the appearance of ALRMOUT at the output even if ALRMOUT_PIN =1.
Once an LOS or FOS alarm is asserted for one of the input clocks, it is held high until the input clock is validated
over a designated time period. The validation time is programmable via the VALTIME[1:0] register bits as shown in
Table 8, “AC Characteristics—All Devices”. If another error condition on the same input clock is detected during the
validation time then the alarm remains asserted and the validation time starts over.
Note that hitless switching between input clocks applies only when the input clock validation time
VALTIME[1:0] = 01 or higher.
Table 53. Alarm Output Logic Equations [Si5368 and CKCONFIG_REG =1]
FOS_ENAlarm Output Equations
0
(Disables FOS)
1C1B = LOS1_INT or (LOS3_INT and FSYNC_SWTCH_REG) or
7.11.7. LOS Algorithm for Reference Clock Input (Si5319, Si5324, Si5326, Si5327, Si5368, Si5369 , Si5374,
Si5375)
The reference clock input on the XA/XB port is monitored for LOS. The LOS circuitry d ivides the signal at XA/XB by
128, producing a 78 kHz to 1.2 MHz signal, and monitors the signal for LOS using the same algorithm as described
in Section “7.11.1. Loss-of-Signal Alarms (Si5319, Si5324, Si5325, Si5326, Si5327, Si5367, Si5368, Si5369,
Si5374, Si5375)”. The LOSX_INT read only bit reflects the state of a loss-of-signal monitor on the XA/XB port. For
the Si5374 and Si5375, the XA/XB port refers to the OSC_P and OSC_N pins.
The device has a PLL lock detection algorithm that indicates the lock status on the LOL output pin and the
LOL_INT read-only register bit. The algorithm works by continuously monitoring the phase of the input clock in
relation to the phase of the feedbac k clock. A retriggerable one-sho t is set each time a potential phase cy cle slip
condition is detected. If no potential phase cycle slip occurs for the retrigger time, the LOL output is set low,
indicating the PLL is in lock. The LOL pin is held in the active state during an internal PLL calibration. The active
polarity of the LOL output pin is set using the LOL_POL register bit (default active high).
The lock detect retrigger time is user-selectable, independent of the loop bandwidth. The LOCKT[2:0] register bits
must be set by the user to the desired setting. Table 54 shows the lock detect retrigger time for both modes of
operation. LOCKT is the minimum amount of time that LOL will be active.
C1B = LOS1_INT or (LOS3_INT and FSYNC_SWTCH_REG)
C2B = LOS2_INT or (LOS4_INT and FSYNC_SWTCH_REG)
C3B tri-state,
ALRMOUT = ALIGN_INT
FOS1_INT
C2B = LOS2_INT or (LOS4_INT and FSYNC_SWTCH_REG) or
FOS2_INT
C3B tri-state,
ALRMOUT = ALIGN_INT
100Rev. 0.5
Loading...
+ hidden pages
You need points to download manuals.
1 point = 1 manual.
You can buy points or you can get point for every manual you upload.