Continuous Rate 8.5 Gbps to 11.3 Gbps Clock and
Data Recovery IC with Integrated Limiting Amp/EQ
Data Sheet
ADN2917
FEATURES
Serial data input: 8.5 Gbps to 11.3 Gbps
No reference clock required
Exceeds SONET/SDH requirements for jitter
transfer/generation/tolerance
Quantizer sensitivity: 9.2 mV p-p typical (limiting amplifier
mode)
Optional limiting amplifier and equalizer inputs
Programmable jitter transfer bandwidth to support G.8251 OTN
Programmable slice level
Sample phase adjust
Output polarity invert
Programmable LOS threshold via I2C
I2C to access optional features
LOS alarm (limiting amplifier mode only)
LOL indicator
PRBS generator/detector
Application-aware power
352 mW at 8.5 Gbps, equalizer mode, no clock output
430 mW at 11.3 Gbps, equalizer mode, no clock output
Power supplies: 1.2 V, flexible 1.8 V to 3.3 V, and 3.3 V
4 mm × 4 mm 24-lead LFCSP
APPLICATIONS
SONET/SDH OC-192, 10GFC, and 10GE and all associated FECs
XFP, line cards, clocks, routers, repeaters, instruments
Any rate regenerators/repeaters
GENERAL DESCRIPTION
The ADN2917 provides the receiver functions of quantization,
signal level detect, and clock and data recovery for continuous
data rates from 8.5 Gbps to 11.3 Gbps. The ADN2917 automati-
cally locks to all data rates without the need for an external
reference clock or programming. ADN2917 jitter performance
exceeds all jitter specifications required by SONET/SDH,
including jitter transfer, jitter generation, and jitter tolerance.
The ADN2917 provides manual or automatic slice adjust and
manual sample phase adjusts. Additionally, the user can select a
limiting amplifier or equalizer at the input. The equalizer is
either adaptive or can be manually set.
The receiver front-end loss of signal (LOS) detector circuit
indicates when the input signal level has fallen below a user-
programmable threshold. The LOS detect circuit has hysteresis
to prevent chatter at the LOS output. In addition, the input
signal strength can be read through the I2C registers.
The ADN2917 also supports pseudorandom binary sequence
(PRBS) generation, bit error detection, and input data rate
readback features.
The ADN2917 is available in a compact 4 mm × 4 mm, 24-lead
frame chip scale package (LFCSP). All ADN2917 specifications
are defined over the ambient temperature range of −40°C to
+85°C, unless otherwise noted.
FUNCTIONAL BLOCK DIAGRAM
LOS
THRESH
LOS
DETECT
I
2
C REGISTERS
LA
BYPASS
V
CC
FLOAT
V
CM
I
2
C
I
2
C
CML CML
EQ
DATA
INPUT
SAMPLER
SAMPLE
PHASE
ADJUST
RXD
TXD
CLK
DDR
RXCK
FREQUENCY
ACQUISITION
AND LOCK
DETECTOR
FIFO
DOWNSAMPLER
AND LOOP
FILTER
PHASE
SHIFTER
DCO
CLOCK
÷2÷N
LOS
PIN
NIN
SCK SDA LOL
DATA RATE
REFCLKP/
REFCLKN
(OPTIONAL) DATOUTP/
DATOUTN CLKOUTP/
CLKOUTN
11778-001
ADN2917
2
50Ω50Ω
SLICE
ADJUST
I
2
C_ADDR
Figure 1.
Rev. 0 Document Feedback
Information furnished by Analog Devices is believed to be accurate and reliable. However, no
responsibility is assumed by Analog Devices for its use, nor for any infringements of patents or other
rights of third parties that may result from its use. Specifications subject to change without notice. No
license is granted by implication or otherwise under any patent or patent rights of Analog Devices.
Trademarks and registered trademarks are the property of their respective owners.
One Technology Way, P.O. Box 9106, Norwood, MA 02062-9106, U.S.A.
Tel: 781.329.4700 ©2014 Analog Devices, Inc. All rights reserved.
Technical Support www.analog.com
ADN2917 Data Sheet
TABLE OF CONTENTS
Features .............................................................................................. 1
Applications ....................................................................................... 1
General Description ......................................................................... 1
Functional Block Diagram .............................................................. 1
Revision History ............................................................................... 2
Specifications ..................................................................................... 3
Jitter Specifications ....................................................................... 4
Output and Timing Specifications ............................................. 5
Timing Diagrams .......................................................................... 7
Absolute Maximum Ratings ............................................................ 8
Thermal Characteristics .............................................................. 8
ESD Caution .................................................................................. 8
Pin Configuration and Function Descriptions ............................. 9
Typical Performance Characteristics ........................................... 10
I2C Interface Timing and Internal Register Descriptions ......... 12
Register Map ............................................................................... 13
Theory of Operation ...................................................................... 18
Functional Description .................................................................. 20
Frequency Acquisition ............................................................... 20
Limiting Amplifier ..................................................................... 20
Slice Adjust .................................................................................. 20
Edge Select ................................................................................... 20
Loss of Signal Detector .............................................................. 22
Passive Equalizer ........................................................................ 22
0 dB EQ ........................................................................................ 23
Lock Detector Operation .......................................................... 23
Output Disable and Squelch ..................................................... 24
I2C Interface ................................................................................ 24
Reference Clock (Optional) ...................................................... 24
Additional Features Available via the I2C Interface ............... 26
Input Configurations ................................................................. 28
DC-Coupled Application .......................................................... 31
Outline Dimensions ....................................................................... 32
Ordering Guide .......................................................................... 32
REVISION HISTORY
5/14—Revision 0: Initial Version
Rev. 0 | Page 2 of 32
Data Sheet ADN2917
SPECIFICATIONS
TA = TMIN to TMAX, VCC = VCCMIN to VCCMAX, VCC1 = VCC1MIN to VCC1MAX, VDD = VDDMIN to VDDMAX, VEE = 0 V, input data
pattern: PRBS 223 − 1, ac-coupled, I2C register default settings, unless otherwise noted.
Table 1.
Parameter Test Conditions/Comments Min Typ Max Unit
DATA RATE SUPPORT RANGE
8.5
11.3
Gbps
INPUT—DC CHARACTERISTICS
Peak-to-Peak Differential Input1 PIN – NIN 1.0 V
Input Resistance Differential 95 100 105 Ω
BYPASS PATH—CML INPUT
Input Voltage Range At PIN or NIN, dc-coupled, RX_TERM_FLOAT = 1 (float) 0.5 VCC V
Input Common-Mode Level DC-coupled (see Figure 32), 600 mV p-p differential,
RX_TERM_FLOAT = 1 (float)
0.65 VCC − 0.15 V
Differential Input Sensitivity
OC-192 AC-coupled, RX_TERM_FLOAT = 0 (VCM = 1.2 V), bit
error rate (BER) = 1 × 10−10
200 mV p-p
8GFC
2
Jitter tolerance scrambled pattern (JTSPAT ),
ac-coupled, RX_TERM_FLOAT = 0 (VCM = 1.2 V),
BER = 1 × 10−12
200 mV p-p
LIMITING AMPLIFIER INPUT PATH
Differential Input Sensitivity
OC-192 BER = 1 × 10−10 9.2 mV p-p
8GFC2 JTSPAT, BER = 1 × 10−12 8.3 mV p-p
10.3125 Gbps
JTSPAT, BER = 1 × 10−12
11.0
mV p-p
EQUALIZER INPUT PATH
Differential Input Sensitivity 15-inch FR-4, 100 Ω differential transmission line,
adaptive EQ on
8GFC
2
JTSPAT, BER = 1 × 10
−12
115 mV p-p
OC-192 BER = 1 × 10
−10
184 mV p-p
INPUT—AC CHARACTERISTICS
S11 At 7.5 GHz, differential return loss, see Figure 9 −12 dB
LOS DETECT
Loss of Signal Detect 10 mV p-p
Loss of signal minimum program value 5 mV p-p
Loss of signal maximum program value 128 mV p-p
Hysteresis (Electrical) 5.7 dB
LOS Assert Time AC-coupled3 135 µs
LOS Deassert Time AC-coupled
3
110 µs
LOSS OF LOCK (LOL) DETECT
DCO Frequency Error for LOL Assert With respect to nominal, data collected in lock to
reference (LTR) mode
1000 ppm
DCO Frequency Error for LOL Deassert With respect to nominal, data collected in LTR mode 250 ppm
LOL Assert Response Time 8.5 Gbps, JTSPAT 25 µs
10 Gbps 18 µs
ACQUISITION TIME
Lock to Data (LTD) Mode OC192 0.5 ms
11.3 Gbps 0.5 ms
8.5 Gbps, JTSPAT 0.5 ms
Optional LTR Mode4
6.0
ms
DATA RATE READBACK ACCURACY
Coarse Readback ±5 %
Fine Readback In addition to reference clock accuracy ±100 ppm
Rev. 0 | Page 3 of 32
ADN2917 Data Sheet
Parameter Test Conditions/Comments Min Typ Max Unit
POWER SUPPLY VOLTAGE
VCC 1.14 1.2 1.26 V
VDD 2.97 3.3 3.63 V
VCC1 1.62 1.8 3.63 V
POWER SUPPLY CURRENT Limiting amplifier mode, clock output enabled
VCC 8GFC,2 JTSPAT 319.1 359.5 mA
OC-192 333 377.4 mA
VDD 8GFC,
2
JTSPAT 7.20 8.1 mA
OC-192 7.21 8.59 mA
VCC1 8GFC,
2
JTSPAT 22.2 28.4 mA
OC-192 35.1 47.4 mA
TOTAL POWER DISSIPATION
Clock Output Enabled Limiting amplifier mode, 8.5 Gbps 446.6 mW
Limiting amplifier mode, 9.953 Gbps 486.5 mW
Clock Output Disabled Equalizer mode, 8.5 Gbps 352 mW
Equalizer mode, 11.3 Gbps
430
mW
OPERATING TEMPERATURE RANGE −40 +85 °C
1 See Figure 33.
2 Fibre Channel Physical Interface 4 standard, FC-PI-4, Rev 8.00, May 21, 2008.
3 When ac-coupled, the LOS assert and deassert times are dominated by the RC time constant of the ac coupling capacitor and the 100 Ω differential input termination
of the ADN2917 input stage.
4 This typical acquisition specification applies to all selectable reference clock frequencies in the range of 11.05 MHz to 176.8 MHz.
JITTER SPECIFICATIONS
TA = TMIN to TMAX, VCC = VCCMIN to VCCMAX, VCC1 = VCC1MIN to VCC1MAX, VDD = VDDMIN to VDDMAX, VEE = 0 V, input data
pattern: PRBS 223 − 1, ac-coupled to 100 Ω differential termination load, I2C register default settings, unless otherwise noted.
Table 2.
Parameter Test Conditions/Comments Min Typ Max Unit
PHASE-LOCKED LOOP
CHARACTERISTICS
Jitter Transfer Bandwidth (BW)
1
OC-192 TRANBW[2:0] = 3 1064 1650 kHz
OTN mode,
2
TRANBW[2:0] = 1 294 529 kHz
8GFC3 1242 1676 kHz
Jitter Peaking
OC-192 20 kHz to 80 MHz 0.014 0.024 dB
8GFC
3
20 kHz to 80 MHz 0.004 0.021 dB
Jitter Generation
OC-192 Unfiltered 0.0045 0.0067 UI rms
Unfiltered 0.076 UI p-p
8GFC
3
Unfiltered
0.005
UI rms
Unfiltered
0.044
UI p-p
Jitter Tolerance TRANBW[2:0] = 4 (default)
OC-192 2000 Hz 4255 UI p-p
20 kHz 106 UI p-p
400 kHz 3.78 UI p-p
4 MHz 0.36 0.50 UI p-p
80 MHz 0.28 0.43 UI p-p
Rev. 0 | Page 4 of 32
Data Sheet ADN2917
Parameter Test Conditions/Comments Min Typ Max Unit
8GFC,
3
JTSPAT
Sinusoidal Jitter at 340 kHz 6.7 UI p-p
Sinusoidal Jitter at 5.098 MHz 0.53 UI p-p
Sinusoidal Jitter at 80 MHz 0.59 UI p-p
Rx Jitter Tracking Test4
Voltage modulation amplitude (VMA) = 170 mV p-p at
100 MHz, 425 mV p-p at 100 MHz, 170 mV p-p at 2.5 GHz,
and 425 mV p-p at 2.5 GHz excitation frequency5
510 kHz, 1 UI 10−12 <10−12 BER
100 kHz, 5 UI 10−12 <10−12 BER
1 Jitter transfer bandwidth is programmable by adjusting TRANBW[2:0] in the DPLLA register (Register 0x10).
2 Set TRANBW[2:0] (Bits[D2:D0] in Register 0x10) = 1 to enter OTN mode. OTN is the optical transport network as defined in ITU G.709.
3 Fibre Channel Physical Interface 4 standard, FC-PI-4, Rev 8.00, May 21, 2008.
4 Conditions of FC-PI-4, Rev 8.00, Table 27, 800-DF-EL-S apply.
5 Must have zero errors during the tests for an interval of time that is ≤10−12 BER to pass the tests.
OUTPUT AND TIMING SPECIFICATIONS
TA = TMIN to TMAX, VCC = VCCMIN to VCCMAX, VCC1 = VCC1MIN to VCC1MAX, VDD = VDDMIN to VDDMAX, VEE = 0 V, input data
pattern: PRBS 223 − 1, ac-coupled to 100 Ω differential termination load, I2C register default settings, unless otherwise noted.
Table 3.
Parameter Test Conditions/Comments Min Typ Max Unit
CML OUTPUT CHARACTERISTICS
Data Differential Output Swing OC-192, DATA_SWING[3:0] (Bits[D7:D4] in
Register 0x1F) setting = 0xC (default)
535 600 672 mV p-p
OC-192, DATA_SWING[3:0] setting = 0xF (maximum) 668 724 771 mV p-p
OC-192, DATA_SWING[3:0] setting = 0x4 (minimum) 189 219 252 mV p-p
Clock Differential Output Swing OC-192, CLOCK_SWING[3:0] (Bits[D3:D0] in
Register 0x1F) setting = 0xC (default)
406 508 570 mV p-p
OC-192, CLOCK_SWING[3:0] setting = 0xF (maximum) 448 583 659 mV p-p
OC-192, CLOCK_SWING[3:0] setting = 0x4 (minimum) 162 217 249 mV p-p
Data Differential Output Swing 8GFC, DATA_SWING[3:0] setting = 0xC (default) 540 600 666 mV p-p
8GFC, DATA_SWING[3:0] setting = 0xF (maximum) 662 725 778 mV p-p
8GFC, DATA_SWING[3:0] setting = 0x4 (minimum 190 214 245 mV p-p
Clock Differential Output Swing 8GFC, CLOCK_SWING[3:0] setting = 0xC (default) 426 518 588 mV p-p
8GFC, CLOCK_SWING[3:0] setting = 0xF (maximum) 489 603 680 mV p-p
8GFC, CLOCK_SWING[3:0] setting = 0x4 (minimum) 166 213 245 mV p-p
Output High Voltage VOH, dc-coupled VCC – 0.05 VCC −
0.025
VCC V
Output Low Voltage VOL, dc-coupled VCC – 0.36 VCC −
0.325
VCC −
0.29
V
CML OUTPUT TIMING CHARACTERISTICS
Rise Time 20% to 80%, at OC-192, DATOUTN/DATOUTP 17.4 32.6 46.5 ps
20% to 80%, at OC-192, CLKOUTN/CLKOUTP 22.2 28.3 33.1 ps
20% to 80%, at 8GFC,
1
DATOUTN/DATOUTP 20.4 33.1 44 ps
20% to 80%, at 8GFC,1 CLKOUTN/CLKOUTP 23.1 29.7 35.8 ps
Fall Time 80% to 20%, at OC-192, DATOUTN/DATOUTP 17.5 33 49.1 ps
80% to 20%, at OC-192, CLKOUTN/CLKOUTP
23.9
29.2
33.7
ps
80% to 20%, at 8GFC,1 DATOUTN/DATOUTP 23 34.2 46.8 ps
80% to 20%, at 8GFC,
1
CLKOUTN/CLKOUTP 25 31.3 37.1 ps
Setup Time, Full Rate Clock t
S
(see Figure 2) 0.5 UI
Hold Time, Full Rate Clock tH (see Figure 2) 0.5 UI
Setup Time, DDR Clock t
S
(see Figure 3) 0.5 UI
Hold Time, DDR clock tH (see Figure 3) 0.5 UI
Rev. 0 | Page 5 of 32
ADN2917 Data Sheet
Parameter Test Conditions/Comments Min Typ Max Unit
I2C INTERFACE DC CHARACTERISTICS LVTTL
Input High Voltage V
IH
2.0 V
Input Low Voltage V
IL
0.8 V
Input Current VIN = 0.1 × VDD or VIN = 0.9 × VDD −10.0 +10.0 µA
Output Low Voltage V
OL
, I
OL
= 3.0 mA 0.4 V
I
2
C INTERFACE TIMING See Figure 17
SCK Clock Frequency 400 kHz
SCK Pulse Width High tHIGH 600 ns
SCK Pulse Width Low t
LOW
1300 ns
Start Condition Hold Time tHD;STA 600 ns
Start Condition Setup Time t
SU;STA
600 ns
Data Setup Time t
SU;DAT
100 ns
Data Hold Time tHD;DAT 300 ns
SCK/SDA Rise/Fall Time
2
t
R
/t
F
20 + 0.1 C
b
300 ns
Stop Condition Setup Time tSU;STO 600 ns
Bus Free Time Between Stop and
Start Conditions
tBUF 1300 ns
LVTTL DC INPUT CHARACTERISITICS
(I2C_ADDR)
Input Voltage
High
VIH
2.0
V
Low VIL 0.8 V
Input Current
High I
IH
, V
IN
= 2.4 V +5 µA
Low IIL, VIN = 0.4 V −5 µA
LVTTL DC OUTPUT CHARACTERISITICS
(LOS/LOL)
Output Voltage
High V
OH
, I
OH
= 2.0 mA 2.4 V
Low V
OL
, I
OL
= −2.0 mA 0.4 V
REFERENCE CLOCK CHARACTERISTICS Optional LTR mode
Input Compliance Voltage (Single-
Ended)
VCM (no input offset, no input current),
see Figure 25, ac-coupled input
0.55 1.0 V
Minimum Input Drive See Figure 25, ac-coupled, differential input 100 mV p-p diff
Reference Frequency 11.05 176.8 MHz
Required Accuracy3 AC-coupled, differential input 100 ppm
1 Fibre Channel Physical Interface 4 standard, FC-PI-4, Rev 8.00, May 21, 2008.
2 Cb is the total capacitance of one bus line in picofarads (pF). If mixed with high speed (HS) mode devices, faster rise/fall times are allowed (refer to the Philips I2C Bus
Specification, Version 2.1).
3 Required accuracy in dc-coupled mode is guaranteed by design as long as the clock common-mode voltage output matches the reference clock common-mode
voltage range.
Rev. 0 | Page 6 of 32
Data Sheet ADN2917
Rev. 0 | Page 7 of 32
TIMING DIAGRAMS
11778-002
CLKOUTP
DATOUTP/
DATOUTN
t
S
t
H
Figure 2. Data to Clock Timing (Full Rate Clock Mode)
11778-003
CLKOUTP
DATOUTP/
DATOUTN
t
H
t
S
Figure 3. Data to Clock Timing (Half Rate Clock/DDR Mode)
11778-004
D
A
TOUTP
LOGIC HIGH LOGIC LOW LOGIC HIGH LOGIC LOW
LOGIC HIGH LOGIC LOW LOGIC HIGH LOGIC LOW
DATOUTN
DATOUTP – D ATOUT N
0V V
SE
V
SE
V
DIFF
Figure 4. Single-Ended vs. Differential Output Amplitude Relationship
ADN2917 Data Sheet
ABSOLUTE MAXIMUM RATINGS
Table 4.
Parameter Rating
Supply Voltage (VCC = 1.2 V) 1.26 V
Supply Voltage (VDD and VCC1 = 3.3 V)
3.63 V
Maximum Input Voltage (REFCLKP/REFCLKN,
NIN/PIN)
1.26 V
Minimum Input Voltage (REFCLKP/REFCLKN,
NIN/PIN)
VEE – 0.4 V
Maximum Input Voltage (SDA, SCK,
I2C_ADDR)
3.63 V
Minimum Input Voltage (SDA, SCK,
I2C_ADDR)
VEE – 0.4 V
Maximum Junction Temperature 125°C
Storage Temperature Range −65°C to +150°C
Lead Temperature (Soldering, 10 sec)
300°C
Stresses at or above those listed under Absolute Maximum
Ratings may cause permanent damage to the product. This is a
stress rating only; functional operation of the product at these
or any other conditions above those indicated in the operational
section of this specification is not implied. Operation beyond
the maximum operating conditions for extended periods may
affect product reliability.
THERMAL CHARACTERISTICS
Thermal Resistance
Thermal resistance is specified for the worst-case conditions,
that is, a device soldered in a circuit board for surface-mount
packages, for a 4-layer board with the exposed paddle soldered
to VEE.
Table 5. Thermal Resistance
Package Type θ
JA
1 θ
JB
2 θ
JC
3 Unit
24-Lead LFCSP_VQ 45 5 11 °C/W
1 Junction to ambient.
2 Junction to base.
3 Junction to case.
ESD CAUTION
Rev. 0 | Page 8 of 32
Data Sheet ADN2917
PIN CONFIGURATION AND FUNCTION DESCRIPTIONS
NOTES
1. DNC = DO NO T CO NNECT.
2. THE EXPOSED PAD ON THE BOTTOM OF THE DEVICE
PACKAGE M US T BE CONNECTED T O VE E E LECT RICAL LY.
THE E X P OSE D P AD WORKS AS A HE AT SINK.
PIN 1
INDICATOR
1VCC 2PIN 3NIN 4VEE 5LOS 6LOL
VEE
VCC1
VDD
CLKOUTN
CLKOUTP
VEE
15
16
17
18
14
13
DATOUTP
DNC
VDD
VCC
DATOUTN
VCC
VCC
REFCLKN
REFCLKP
SDA
SCK
7
8
9
1121
01 12223242
0291
ADN2917
TOP VIEW
(Not to
Scale)
11778-005
I
2
C_ADDR
Figure 5. Pin Configuration
Table 6. Pin Function Descriptions
Pin No.
Mnemonic
Type 1
Description
1 VCC P 1.2 V Supply for Limiting Amplifier.
2 PIN AI Positive Differential Data Input (CML).
3 NIN AI Negative Differential Data Input (CML).
4 VEE P Ground for Limiting Amplifier.
5 LOS DO Loss of Signal Output (Active High).
6 LOL DO Loss of Lock Output (Active High).
7 VEE P Digital Control Oscillator (DCO) Ground.
8
VCC1
P
1.8 V to 3.3 V DCO Supply.
9 VDD P 3.3 V High Supply.
10 CLKOUTN DO Negative Differential Recovered Clock Output (CML).
11 CLKOUTP DO Positive Differential Recovered Clock Output (CML).
12 VEE P Ground for CML Output Drivers.
13 VCC P 1.2 V Supply for CML Output Drivers.
14 DATOUTN DO Negative Differential Retimed Data Output (CML).
15 DATOUTP DO Positive Differential Retimed Data Output (CML).
16 DNC DI Do Not Connect. Tie off to ground. Leave this pin floating.
17 VDD P 3.3 V High Supply.
18 VCC P 1.2 V Core Digital Supply.
19 SCK DI Clock for I2C.
20 SDA DIO Bidirectional Data for I
2
C.
21 VCC P 1.2 V Core Supply.
22 I2C_ADDR DI I2C Address. Sets the device I2C address = 0x80 when I2C_ADDR = 0, and the device I2C address = 0x82
when I2C_ADDR = 1.
23 REFCLKN DI Negative Reference Clock Input (Optional).
24 REFCLKP DI Positive Reference Clock Input (Optional).
EPAD P Exposed Pad (VEE). The exposed pad on the bottom of the device package must be connected to VEE
electrically. The exposed pad works as a heat sink.
1 P is power, AI is analog input, DI is digital input, DO is digital output, and DIO is digital input/output.
Rev. 0 | Page 9 of 32
ADN2917 Data Sheet
TYPICAL PERFORMANCE CHARACTERISTICS
TA = 25°C, VCC = 1.2 V, VCC1 = 1.8 V, VDD = 3.3 V, VEE = 0 V, input data pattern: PRBS 215 − 1, ac-coupled inputs and outputs,
unless otherwise noted.
11778-006
16.8ps/DIV
61.4mV/DIV
Figure 6. Output Eye Diagram at OC-192
5
–45
–15
–5
–20
–25
–30
–35
–40
–10
0
1k 10k 100k 1M 10M 100M
11778-007
JIT TER TRANSFER (dB)
FRE Q UE NCY ( Hz )
XFP MASK
Figure 7. Jitter Transfer: OC-192,
TRANBW[2:0] (Bits[D2:D0] in Register 0x10) = 3
1k
0.01
0.1
1
10
100
100 1k 10k 100k 1M 10M 100M
11778-008
JITTER AMPLITUDE (UI)
JIT TER FREQUENCY ( Hz )
ADN2917 TOLE RANCE
SONET REQUIREMENT MASK
Figure 8. Jitter Tolerance: OC-192
0
–40
–15
–5
–20
–25
–30
–35
–10
1M 10M 100M 1G 10G 100G
11778-009
LOG MAGNITUDE (dB)
FRE Q UE NCY ( Hz )
Figure 9. Typical S11 Spectrum Performance
Rev. 0 | Page 10 of 32
Data Sheet ADN2917
0.6
0.5
0.4
0.3
0.2
0.1
00161412108642
11778-011
BER
EQ SETTING
TYPICAL
ADAPTIVE EQ
SETTING
Figure 10. BER in Equalizer Mode vs. EQ Compensation at OC-192
(Measured with an OC-192 Signal of 400 mV p-p diff, on 15-Inch FR4 Traces,
with Variant EQ Compensation, Including Adaptive EQ)
12
0
2
4
6
8
10
11778-010
SENSITIVITY (mV p-p DIFF)
DATA RATE (Gb ps)
2.66600 × 10
9
9.95328 × 10
9
1.07090 × 10
10
Figure 11. Sensitivities of SONET/SDH Data Rates (BER = 10−10)
16
0
2
4
6
8
12
14
10
11778-012
SENSITIVITY (mV p-p DIFF)
DATA RATE (Gb ps)
8.50000 × 10
9
9.95328 × 10
9
1.03125 × 10
10
1.10957 × 10
10
Figure 12. Sensitivities of Non-SONET/SDH Data Rates (BER = 10−12)
Rev. 0 | Page 11 of 32
ADN2917 Data Sheet
I2C INTERFACE TIMING AND INTERNAL REGISTER DESCRIPTIONS
11778-013
1
SLAVE ADDRESS[6:0]
SET BY
PIN 22
MSB = 1 0 = W
1 = R
R/W
CTRL.
00000xx
Figure 13. Slave Address Configuration
11778-014
S SLAVE ADDR, LS B = 0 ( W) A(S) A(S) A(S)
DATASUBADDR A(S) P
DATA
Figure 14. I2C Write Data Transfer
11778-015
S
S = START BIT P = STOP BIT
A(S) = ACKNOW LEDG E BY S LAVE A(M) = ACKNOWLEDGE BY MASTER A(M) = NO ACKNOWL EDGE BY M AS TER
SSLAV E ADDR, L S B = 0 ( W) SLAV E ADDR, L S B = 1 ( R)A(S) A(S)SUBADDR A(S) DATA A(M) DATA PA(M)
Figure 15. I2C Read Data Transfer
11778-016
START BIT
S
STOP BIT
P
ACK
ACK
WR ACK
D0
D7
A0
A7
A5
A6
SLAVE ADDR[4:0]
SLAVE ADDRESS SUBADDRESS DATA
SUBADDR[6:1] DATA[6:1]
SCK
SDA
Figure 16. I2C Data Transfer Timing Diagram
11778-017
tBUF
SDA
S S P S
SCK
tF
tLOW
tRtF
tHD;STA
tHD;DAT
tSU;DAT
tHIGH tSU;STA
tSU;STO
tHD;STA
tR
Figure 17. I2C Interface Timing Diagram
Rev. 0 | Page 12 of 32
Data Sheet ADN2917
REGISTER MAP
Writing to register bits other than those clearly labeled is not recommended and may cause unintended results.
Table 7. Internal Register Map
Reg Name R/W
Addr
(Hex)
Default
(Hex) D7 D6 D5 D4 D3 D2 D1 D0
Readback/Status
FREQMEAS0 R 0x0 N/A FREQ0[7:0] (RATE_FREQ[7:0])
FREQMEAS1 R 0x1 N/A FREQ1[7:0] (RATE_FREQ[15:8])
FREQMEAS2 R 0x2 N/A FREQ2[7:0] (RATE_FREQ[23:16])
FREQ_RB1 R 0x4 N/A VCOSEL[7:0]
FREQ_RB2 R 0x5 N/A FULLRATE DIVRATE[3:0] VCOSEL[9:8]
STATUSA R 0x6 N/A LOS
status
LOL status LOS done Static LOL RATE_
MEAS_
COMP
General Control
CTRLA R/W 0x8 0x10 0 CDR_MODE[2:0] 0 Reset static
LOL
RATE_
MEAS_
EN
RATE_MEAS_
RESET
CTRLB R/W 0x9 0x00 SOFTWARE_
RESET
INIT_
FREQ_
ACQ
0 LOL_
CONFIG
LOS PDN LOS polarity 0 0
CTRLC R/W 0xA 0x04 0 0 0 0 0 REFCLK_
PDN
0 1
FLL Control
LTR_MODE R/W 0xF 0x00 0 LOL data FREF_RANGE[1:0] DATA_TO_REF_RATIO[3:0]
D/PLL Control
DPLLA R/W 0x10 0x1C 0 0 0 EDGE_SEL[1:0] TRANBW[2:0]
DPLLD R/W 0x13 0x06 0 0 0 0 0 ADAPTIVE_
SLICE_EN
DLL_SLEW[1:0]
Phase R/W 0x14 0x00 0 0 0 0 SAMPLE_PHASE[3:0]
Slice W 0x15 N/A Extended
slice
Slice[6:0]
LA_EQ R/W 0x16 0x08 RX_
TERM_
FLOAT
INPUT_SEL[1:0] ADAPTIVE_
EQ_EN
EQ_BOOST[3:0]
Slice
Readback
R 0x73 N/A SLICE_RB[7:0]
Output Control
OUTPUTA R/W 0x1E 0x00 0 0 Data
squelch
DATOUT_
DISABLE
CLKOUT_
DISABLE
0 DATA_
POLARITY
CLOCK_
POLARITY
OUTPUTB R/W 0x1F 0xCC DATA_SWING[3:0] CLOCK_SWING[3:0]
LOS Control
LOS_DATA R/W 0x36 0x00 LOS_DATA[7:0]
LOS_THRESH R/W 0x38 0x0A LOS_THRESHOLD[7:0]
LOS_CTRL
R/W
0x74
0x00
0
0
LOS_
WRITE
LOS_
ENABLE
LOS_
RESET
LOS_ADDRESS[2:0]
PRBS Control
PRBS Gen 1 R/W 0x39 0x00 0 0
DATA_
CID_
BIT
DATA_
CID_
EN
0
DATA_GEN_
EN
DATA_GEN_MODE[1:0]
PRBS Gen 2 R/W 0x3A 0x00 DATA_CID_LENGTH[7:0]
PRBS Gen 3 R/W 0x3B 0x00 PROG_DATA[7:0]
PRBS Gen 4 R/W 0x3C 0x00 PROG_DATA[15:8]
PRBS Gen 5 R/W 0x3D 0x00 PROG_DATA[23:16]
PRBS Gen 6
R/W
0x3E
0x00
PROG_DATA[31:24]
PRBS Rec 1 R/W 0x3F 0x00 0 0 0 0 DATA_
RECEIVER_
CLEAR
DATA_
RECEIVER_
ENABLE
DATA_RECEIVER_
MODE[1:0]
PRBS Rec 2 R 0x40 0x00 PRBS_ERROR_COUNT[7:0]
PRBS Rec 3 R 0x41 0x00 PRBS_ERROR
Rev. 0 | Page 13 of 32
ADN2917 Data Sheet
Reg Name R/W
Addr
(Hex)
Default
(Hex) D7 D6 D5 D4 D3 D2 D1 D0
PRBS Rec 4 R 0x42 N/A DATA_LOADED[7:0]
PRBS Rec 5 R 0x43 N/A DATA_LOADED[15:8]
PRBS Rec 6 R 0x44 N/A DATA_LOADED[23:16]
PRBS Rec 7
R
0x45
N/A
DATA_LOADED[31:24]
ID/Revision
REV
R
0x48
0x54
REV[7:0]
ID R 0x49 0x15 ID[7:0]
HI_CODE R 0x20 0xFF Reserved
LO_CODE R 0x21 0xA6 Reserved
Table 8. Status Register, STATUSA (Address 0x6)
Bits Bit Name Bit Description
D5 LOS status 0 = no loss of signal
1 = loss of signal
D4
LOL status
0 = locked
1 = frequency acquisition mode
D3 LOS done 0 = LOS action not completed
1 = LOS action completed
D2 Static LOL 0 = no LOL event since last reset
1 = LOL event since last reset; clear by using the static LOL bit, CTRLA[D2]
D0 RATE_MEAS_COMP Rate measurement complete
0 = frequency measurement incomplete
1 = frequency measurement complete; clear by using static the LOL bit, CTRLA[D0]
Table 9. Control Register, CTRLA (Address 0x8)
Bits Bit Name Bit Description
D7 0 Reserved to 0.
D6:D4 CDR_MODE[2:0] CDR modes.
001 = lock to data (LTD).
011 = lock to reference (LTR).
000, 010 = reserved.
D3 0 Reserved to 0.
D2 Reset static LOL Set to 1 to clear static LOL.
D1 RATE_MEAS_EN Fine data rate measurement enable. Set to 1 to initiate a rate measurement.
D0 RATE_MEAS_RESET Rate measurement reset. Set to 1 to clear a rate measurement.
Table 10. Control Register, CTRLB (Address 0x9)
Bits Bit Name Bit Description
D7 SOFTWARE_RESET Software reset. Write a 1 followed by a 0 to reset the device.
D6 INIT_FREQ_ACQ Initiate frequency acquisition. Write a 1 followed by a 0 to initiate a frequency acquisition
(optional).
D5 0 Reserved; CDR is always enabled.
D4 LOL_CONFIG LOL configuration.
0 = normal LOL.
1 = static LOL.
D3 LOS PDN LOS power-down.
0 = normal LOS.
1 = LOS powered down.
D2
LOS polarity
LOS polarity.
0 = active high LOS pin.
1 = active low LOS pin.
D1:D0 0 Reserved to 0.
Rev. 0 | Page 14 of 32
Data Sheet ADN2917
Table 11. Control Register, CTRLC (Address 0xA)
Bits Bit Name Bit Description
D7:D3 0 Reserved to 0.
D2 REFCLK_PDN Reference clock power-down. Write a 0 to enable the reference clock.
D1 0 Reserved to 0.
D0 1 Reserved to 1.
Table 12. Lock to Reference Clock Mode Programming Register, LTR_MODE1 (Address 0xF)
Bits Bit Name Bit Description
D7
0
Reserved to 0
D6 LOL data LOL data
0 = valid recovered clock vs. reference clock during tracking
1 = valid recovered clock vs. data during tracking
D5:D4 FREF_RANGE[1:0] f
REF
range
00 = 11.05 MHz to 22.1 MHz (default)
01 = 22.1 MHz to 44.2 MHz
10 = 44.2 MHz to 88.4 MHz
11 = 88.4 MHz to 176.8 MHz
D3:D0 DATA_TO_REF_RATIO[3:0] Data to reference ratio
0000 = 1/2
0001 = 1
0010 = 2
N = 2
(n − 1)
1010 = 512
1 Where DIV_fREF is the divided down reference referred to the 11.05 MHz to 22.1 MHz band (see the Reference Clock (Optional) section).
Data Rate/2(LTR_MODE[3:0] − 1) = REFCLK/2LTR_MODE[5:4]
Table 13. D/PLL Control Register, DPLLA (Address 0x10)
Bits Bit Name Bit Description
D7:D5 0 Reserved to 0.
D4:D3 EDGE_SEL[1:0] Edge for phase detection. See the Edge Select section for further details.
00 = rising and falling edge data.
01 = rising edge data.
10 = falling edge data.
11 = rising and falling edge data.
D2:D0 TRANBW[2:0] Transfer bandwidth. Scales transfer bandwidth. Default value is 4. See the Transfer Bandwidth
section for further details.
Transfer BW = Default BW × (TRANBW[2:0]/4)
Table 14. D/PLL Control Register, DPLLD (Address 0x13)
Bits Bit Name Bit Description
D7:D3 0 Reserved to 0.
D2 ADAPTIVE_SLICE_EN Adaptive slice enable. 1 = enables automatic slice adjust.
D1:D0 DLL_SLEW[1:0] DLL slew. Sets the BW of the DLL. See the DLL Slew section for further details.
Table 15. Phase Control Register, Phase (Address 0x14)
Bits Bit Name Bit Description
D7:D4 0 Reserved to 0.
D3:D0 SAMPLE_PHASE[3:0] Adjust the phase of the sampling instant for data rates above 5.65 Gbps in steps of 1/32 UI. This
register is in twos complement notation. See the Sample Phase Adjust section for further details.
Rev. 0 | Page 15 of 32
ADN2917 Data Sheet
Table 16. Slice Level Control Register, Slice (Address 0x15)
Bits Bit Name Bit Description
D7 Extended slice Extended slice enable.
0 = normal slice mode.
1 = extended slice mode.
D6:D0 Slice[6:0] Slice. Slice is a digital word that sets the input threshold. See the Slice Adjust section for further
details. When Slice[6:0] = 0000000, the slice function is disabled.
Table 17. Input Stage Programming Register, LA_EQ (Address 0x16)
Bits Bit Name Bit Description
D7 RX_TERM_FLOAT Rx termination float.
0 = termination common-mode driven.
1 = termination common-mode floated.
D6:D5 INPUT_SEL[1:0] Input stage select.
00: limiting amplifier.
01: equalizer.
10: 0 dB buffer.
11: undefined.
D4 ADAPTIVE_EQ_EN Enable adaptive EQ.
0 = manual EQ control.
1 = adaptive EQ enabled.
D3:D0 EQ_BOOST[3:0] Equalizer gain. These bits set the EQ gain. See the Passive Equalizer section for further details.
Table 18. Output Control Register, OUTPUTA (Address 0x1E)
Bits Bit Name Bit Description
D7:D6 0 Reserved to 0
D5 Data squelch Squelch
0 = normal data
1 = squelch data
D4 DATOUT_DISABLE Data output disable
0 = data output enabled
1 = data output disabled
D3 CLKOUT_DISABLE Clock output disable
0 = clock output enabled
1 = clock output disabled
D2 0 Reserved; double data rate is always enabled
D1 DATA_POLARITY Data polarity
0 = normal data polarity
1 = flip data polarity
D0 CLOCK_POLARITY Clock polarity
0 = normal clock polarity
1 = flip clock polarity
Rev. 0 | Page 16 of 32
Data Sheet ADN2917
Table 19. Output Swing Register, OUTPUTB (Address 0x1F)
Bits Bit Name Bit Description
D7:D4 DATA_SWING[3:0] Adjust data output amplitude. Step size is approximately 50 mV differential. Default register
value is 0xC. Typical differential data output amplitudes are
0x1 = invalid.
0x2 = invalid.
0x3 = invalid.
0x4 = 200 mV.
0x5 = 250 mV.
0x6 = 300 mV.
0x7 = 345 mV.
0x8 = 390 mV.
0x9 = 440 mV.
0xA = 485 mV.
0xB = 530 mV.
0xC = 575 mV.
0xD = 610 mV.
0xE = 640 mV.
0xF = 655 mV.
D3:D0 CLOCK_SWING[3:0] Adjust clock output amplitude. Step size is approximately 50 mV differential. Default register
value is 0xC. Typical differential clock output amplitudes are
0x1 = invalid.
0x2 = invalid.
0x3 = invalid.
0x4 = 200 mV.
0x5 = 250 mV.
0x6 = 300 mV.
0x7 = 345 mV.
0x8 = 390 mV.
0x9 = 440 mV.
0xA = 485 mV.
0xB = 530 mV.
0xC = 575 mV.
0xD = 610 mV.
0xE = 640 mV.
0xF = 655 mV.
Rev. 0 | Page 17 of 32
ADN2917 Data Sheet
THEORY OF OPERATION
The ADN2917 implements a clock and data recovery for data
rates between 8.5 Gbps and 11.3 Gbps. A front end is configurable
to either amplify or equalize the nonreturn-to-zero (NRZ) input
waveform to full-scale digital logic levels.
To process a high speed input data, the user can choose either a
high gain limiting amplifier with better than 10 mV sensitivity,
or a high-pass passive equalizer with up to 10 dB of boost at
5 GHz with 600 mV sensitivity.
An on-chip LOS detector works with the high sensitivity limiting
amplifier. The default threshold for the LOS is the sensitivity of
the device, with a maximum threshold level of 128 mV p-p. The
limiting amplifier slice threshold can use a factory trim setting,
a user-defined threshold set by the I2C, or an adjusted level for
the best eye opening at the phase detector.
When the input signal is corrupted due to FR-4 or other
impairments in the printed circuit board (PCB) traces, a passive
equalizer can be one of the signal integrity options. The
equalizer high frequency boost is configurable through the I2C
registers, in place of the factory default settings. A user-enabled
adaptation is included that automatically adjusts the equalizer
to achieve the widest eye opening. The equalizer can be
manually set for any data rate from 8.5 Gbps up to 11.3 Gbps.
When a signal is presented to the clock and data recovery (CDR),
the ADN2917 is a delay-locked and phase-locked loop circuit
for clock recovery and data retiming from an NRZ encoded
data stream. Input data is sampled by a high speed clock. A digital
downsampler accommodates data rates spanning three orders of
magnitude. Downsampled data is applied to a binary phase
detector.
The phase of the input data signal is tracked by two separate feed-
back loops. A high speed delay-locked loop path cascades a
digital integrator with a digitally controlled phase shifter on the
DCO clock to track the high frequency components of jitter. A
separate phase control loop composed of a digital integrator
and DCO tracks the low frequency components of jitter. The
initial frequency of the DCO is set by a third loop that
compares the DCO frequency with the input data frequency.
This third loop also sets the decimation ratio of the digital
downsampler.
The delay-locked loop (DLL) and phase-locked loop (PLL)
together track the phase of the input data. For example, when
the clock lags the input data, the phase detector drives the DCO
to a higher frequency and decreases the delay of the clock
through the phase shifter; both of these actions serve to reduce
the phase error between the clock and data. Because the loop
filter is an integrator, the static phase error is driven to zero.
Another view of the circuit is that the phase shifter implements
the zero required for frequency compensation of a second-order
phase-locked loop, and this zero is placed in the feedback path
and, thus, does not appear in the closed-loop transfer function.
Because this circuit has no zero in the closed-loop transfer, jitter
peaking is eliminated.
The delay-locked and phase-locked loops, together, simultane-
ously provide wideband jitter accommodation and narrow-band
jitter filtering. The simplified block diagram in Figure 18 shows
that Z(s)/X(s) is a second-order low-pass jitter transfer function
that provides excellent filtering. The low frequency pole is
formed by dividing the gain of the PLL by the gain of the DLL,
where the upsampling and zero-order hold in the DLL has a gain
approaching N at the transfer bandwidth of the loop. Note that
the jitter transfer has no zero, unlike an ordinary second-order
phase-locked loop. This means that the main PLL loop has no
jitter peaking. This makes the circuit ideal for signal regenerator
applications, where jitter peaking in a cascade of regenerators
can contribute to hazardous jitter accumulation.
The error transfer, e(s)/X(s), has the same high-pass form as an
ordinary phase-locked loop up to the slew rate limit of the DLL
with a binary phase detector. This transfer function is free to be
optimized to give excellent wideband jitter accommodation
because the jitter transfer function, Z(s)/X(s), provides the
narrow-band jitter filtering.
PSH
Z(s) RECOVERED
CLOCK
K
DLL
I – z
–1
K
DCO
s
I – z
–N
I – z
–1
K
PLL
× TRANBW
I – z
–1
N
N
DELAY-LOCKED LOOP (DLL)
PHASE- LO CKE D LO O P (PLL)
ZERO-ORDER HOLD
SAMPLE CLOCK
INPUT
DATA÷N
BINARY
PHASE
DETECTOR
X(s) e(s)
=K
PLL
× TRANBW – K
DCO
s × N × PS H × K
DLL
+ K
PLL
× TRANBW × K
DCO
Z(s)
X(s)
11778-018
Figure 18. CDR Jitter Block Diagram
Rev. 0 | Page 18 of 32
Data Sheet ADN2917
The delay-locked and phase-locked loops contribute to overall
jitter accommodation. At low frequencies of input jitter on the
data signal, the integrator in the loop filter provides high gain to
track large jitter amplitudes with small phase error. In this case,
the oscillator is frequency modulated and jitter is tracked as in
an ordinary phase-locked loop. The amount of low frequency
jitter that can be tracked is a function of the DCO tuning range.
A wider tuning range gives larger accommodation of low fre-
quency jitter. The internal loop control word remains small for
small jitter frequency so that the phase shifter remains close to
the center of its range and, thus, contributes little to the low
frequency jitter accommodation.
At medium jitter frequencies, the gain and tuning range of the
DCO are not large enough to track input jitter. In this case, the
DCO control word becomes large and saturates. As a result, the
DCO frequency dwells at an extreme of its tuning range. The
size of the DCO tuning range, therefore, has only a small effect
on the jitter accommodation. The delay-locked loop control range
is now larger; therefore, the phase shifter takes on the burden of
tracking the input jitter. An infinite range phase shifter is used
on the clock. Consequently, the minimum range of timing
mismatch between the clock at the data sampler and the retiming
clock at the output is limited to 32 UI by the depth of the FIFO.
There are two ways to acquire the data rate. The default mode
frequency locks to the input data, where a finite state machine
extracts frequency measurements from the data to program the
DCO and loop division ratio so that the sampling frequency
matches the data rate to within 250 ppm. The PLL is enabled,
driving this frequency difference to 0 ppm. The second mode is
lock to reference, in which case the user provides a reference
clock between 11.05 MHz and 176.8 MHz. Division ratios must
be written to a serial port register.
Rev. 0 | Page 19 of 32
ADN2917 Data Sheet
FUNCTIONAL DESCRIPTION
FREQUENCY ACQUISITION
The ADN2917 acquires frequency from the data over a range of
data frequencies from 8.5 Gbps to 11.3 Gbps. The lock detector
circuit compares the frequency of the DCO and the frequency
of the incoming data. When these frequencies differ by more
than 1000 ppm, LOL is asserted and a new frequency acquisi-
tion cycle is initiated. The DCO frequency is reset to the bottom
of its range, and the internal division rate is set to its lowest
value of N = 1, which is the highest octave of data rates. The
frequency detector then compares this sampling rate frequency
to the data rate frequency and either increases N by a factor of 2
if the sampling rate frequency is found to be greater than the
data rate frequency, or increases the DCO frequency if the data
rate frequency is found to be greater than the data sampling
rate. Initially, the DCO frequency is incremented in large steps
to aid fast acquisition. As the DCO frequency approaches the
data frequency, the step size is reduced until the DCO frequency is
within 250 ppm of the data frequency, at which point LOL is
deasserted.
When LOL is deasserted, the frequency-locked loop is turned
off. The PLL or DLL pulls in the DCO frequency until the DCO
frequency equals the data frequency.
LIMITING AMPLIFIER
The limiting amplifier has differential inputs (PIN and NIN)
that are each internally terminated with 50 Ω to an on-chip
voltage reference (VCM = 0.95 V typically). The inputs must be ac-
coupled. Input offset is factory trimmed to achieve better than
10 mV p-p typical sensitivity with minimal drift. The limiting
amplifier can be driven differentially or single-ended. DC
coupling of the limiting amplifier is not possible because the
user needs to supply a common-mode voltage to exactly match
the internal common-mode voltage; otherwise, the internal
50 Ω termination resistors absorb the difference in common-
mode voltages.
Another reason the limiting amplifier cannot be dc-coupled is
that the factory trimmed input offset becomes invalid. The
offset is adjusted to zero by differential currents from the slice
adjust digital-to-analog converted (DAC) (see Figure 1). With
ac coupling, all of the current goes to the 50 Ω termination
resistors on the ADN2917. However, with dc coupling, this
current is shared with the external drive circuit, and calibration
of the offset is lost. In addition, the slice adjust must have all the
current from the slice adjust DAC go to the resistors; otherwise,
the calibration is lost (see the Slice Adjust section).
SLICE ADJUST
The quantizer slicing level can be offset by ±100 mV in 1.6 mV
steps or about ±15 mV in 0.24 mV steps to mitigate the effect of
amplified spontaneous emission (ASE) noise or duty cycle
distortion. Quantizer slice adjust level is set by the Slice[6:0]
(Bits[D6:D0] in I2C Register 0x15).
Accurate control of the slice threshold requires the user to read
back the factory trimmed offset, which is stored as a 7-bit
number in the I2C slice readback register (Register 0x73). Use
Table 20 to decode the measured offset of the part, where an
LSB corresponds to 0.24 m V.
Table 20. Program Slice Level, Normal Slice Mode (Extended
Slice = 0)
Slice[6:0] Decimal Value Offset
0000000 0 Slice function disabled
0000001 1 −15 mV
… … …
1000000 64 0 mV
… … …
1111111 127 +14.75 mV
The amount of offset required for manual slice adjust is deter-
mined by subtracting the offset of the device from the desired
slice adjust level. Use Table 20 or Table 21 to determine the code
word to be written to the I2C slice register.
An extended slice with coarser granularity for each LSB step is
found in Table 21. Setting the extended slice bit (Bit 7) = 1 in
Register 0x15 scales the full-scale range of the slice adjust by a
factor of 6.
Table 21. Program Slice Level, Extended Slice Mode (Extended
Slice = 1)
Slice[6:0] Decimal Value Offset
0000000 128 Slice function disabled
0000001 129 −100 mV
… … …
1000000 192 0 mV
… … …
1111111 255 +100 mV
When manual slice is desired, disable the dc offset loop, which
drives duty cycle distortion on the data to 0. Adaptive slice is
disabled by setting ADAPTIVE_SLICE_EN = 0 in the DPLLD
register (Register 0x13).
EDGE SELECT
A binary or Alexander phase detector drives both the DLL and
PLL loops at all division rates. Duty cycle distortion on the
received data leads to a dead band in the phase detector transfer
function if phase errors are measured on both rising and falling
data transitions. This dead band leads to jitter generation of
unknown spectral composition whose peak-to-peak amplitude
is potentially large.
The recommended usage of the device when the dc offset loop
is disabled computes phase errors exclusively on either the
rising data edges with EDGE_SEL[1:0] (Bits[D4:D3] in Register
0x10) = 1 (decimal) or falling data edges with EDGE_SEL[1:0]
(Bits[D4:D3] in Register 0x10) = 2. The alignment of the clock
to the rising data edges with EDGE_SEL[1:0] = 1 is represented
Rev. 0 | Page 20 of 32
Data Sheet ADN2917
by the top two curves in Figure 19. Duty cycle distortion with
narrow 1s moves the significant sampling instance where data is
sampled to the right of center. The alignment of the clock to the
falling data edges with EDGE_SEL[1:0] = 2 is represented by
the first and third curves in Figure 19. The significant sampling
instance moves to the left of center. Sample phase adjust for
rates above 5.65 Gbps can be used to move the significant
sampling instance to the center of the narrow 1 (or narrow 0)
for best jitter tolerance.
EDGE_SEL[1:0]
EDGE_SEL[1:0] = 2
11778-019
CLK1
CLK2
DATA
Figure 19. Phase Detector Timing
DLL Slew
Jitter tolerance beyond the transfer bandwidth of the CDR is
determined by the slew rate of the delay-locked loop implement-
ing a delta modulator on phase. Setting DLL_SLEW[1:0]
(Bits[D1:D0] in Register 0x13) = 2, the default value, configures
the DLL to track 0.75 UI p-p jitter at the highest frequency
breakpoint in the SONET/SDH jitter tolerance mask. This
frequency scales with the rate as fp5 = Rate (Hz)/2500 (for
example, 4.0 MHz for OC-192). Peak-to-peak tracking in UI at
fp4 obeys the expression (1 + DLL_SLEW)/4 UI p-p.
In some applications, full SONET/SDH jitter tolerance is not
needed. In this case, DLL_SLEW[1:0] (Bits[D1:D0] in
Register 0x13) can be set to 0, giving lower jitter generation on
the recovered clock and better high frequency jitter tolerance.
Sample Phase Adjust
The phase of the sampling instant can be adjusted over the I2C
when operating at data rates of 5.65 Gbps or higher by writing to
the SAMPLE_PHASE[3:0] bits (Bits[D3:D0] in Register 0x14).
This feature allows the user to adjust the sampling instant with
the intent of improving the BER and jitter tolerance. Although
the default sampling instant chosen by the CDR is sufficient in
most applications, when dealing with some degraded input
signals, the BER and jitter tolerance performance can be
improved by manually adjusting the phase.
There is a total adjustment range of 0.5 UI, with 0.25 UI in each
direction, in increments of 1/32 UI. SAMPLE_PHASE[3:0]
(Bits[D3:D0] in Register 0x14) is a twos complement number,
and the relationship between data and the sampling clock is
shown in Figure 20.
Transfer Bandwidth
The transfer bandwidth can be adjusted over the I2C by writing to
TRANBW[2:0] (Bits[D2:D0] in Register 0x10). The default value is
4. When set to values below 4, the transfer bandwidth is reduced,
and when set to values above 4, the transfer bandwidth is increased.
The resulting transfer bandwidth is based on the following formula:
×= 4
)( ]TRANBW[2:0
ansfer BWDefault TrWTransfer B
For example, at OC-192, the default transfer bandwidth is
1.4 GHz. The resulting transfer bandwidth when
TRANBW[2:0] is changed is reflected in Table 22.
Table 22. Transfer Bandwidth Adjustments
TRANBW[2:0] Value Transfer BW (kHz)
1
350
2 700
3 1050
4 1400 (default)
5 1750
6 2100
7 2450
Reducing the transfer bandwidth is commonly used in OTN
applications. Never set TRANBW[2:0] = 0, because this makes
the CDR open-loop. Also, note that setting TRANBW[2:0]
above 4 may cause a slight increase in jitter generation and
potential jitter peaking.
11778-021
DATA
CLOCK PHASE = 4
PHASE = 7 PHASE = –4
PHASE = –8
PHASE = 0
(DEFAULT)
Figure 20. Data vs. Sampling Clock LOS Detector Hysteresis
Rev. 0 | Page 21 of 32
ADN2917 Data Sheet
LOSS OF SIGNAL DETECTOR
The receiver front-end LOS detector circuit detects when the
input signal level falls below a user adjustable threshold.
There is typically 6 dB of electrical hysteresis on the LOS
detector to prevent chatter on the LOS pin. This means that, if the
input level drops below the programmed LOS threshold,
causing the LOS pin to assert, the LOS pin is not deasserted
until the input level has increased to 6 dB (2×) above the LOS
threshold (see Figure 21).
11778-020
HYSTERESIS
LOS OUTPUT
INPUT LEVEL
LOS THRESHOLD
t
INPUT VOLTAGE (VDIFF)
Figure 21. LOS Detector Hysteresis
The LOS detector and the slice level adjust can be used simulta-
neously on the ADN2917. Therefore, any offset added to the
input signal by the slice adjust bits (Bits[D6:D0] in Register 0x15)
does not affect the LOS detector measurement of the absolute
input level.
LOS Power-Down
The LOS, by default, is enabled and consumes power. The LOS
is placed in a low power mode by setting LOS PDN (Bit D3 in
Register 0x9) = 1.
LOS Threshold
The LOS threshold has a range between 0 mV and 128 mV and
is set by writing the number of millivolts (mV) to the LOS_DATA
register (Register 0x36), followed by toggling the LOS_ENABLE
bit (Bit D4 in Register 0x74) while LOS_ADDRESS is set to 1.
The following is a procedure for writing the LOS threshold:
1. Write 0x21 to LOS_CTRL (Register 0x74).
2. Write the desired threshold in millivolts to LOS_DATA
(Register 0x36).
3. Write 0x31 to LOS_CTRL (Register 0x74).
4. Write 0x21 to LOS_CTRL (Register 0x74).
The LOS threshold can be set to a value between 0 mV and
63 mV in 1 mV steps and 64 mV to 128 mV in 2 mV steps.
In the lower range, all of the bits are active, giving 1 mV/LSB
resolution, where Bit D0 is the LSB.
However, in the upper range, Bit D0 is disabled (that is, D0 = 0),
making Bit D1 the new LSB and resulting in 2 mV/LSB
resolution.
I2C Register LOS_CTRL (Register 0x74) contains the necessary
address and write enable bits to program this LOS threshold.
Signal Strength Measurement
The LOS measures and digitizes the peak-to-peak amplitude of
the received signal. A single shot measurement is taken by
writing the following sequence of bytes to LOS_CTRL
(Register 0x74). at I2C Address 0x74: 0x7, 0x17, 0x7. Upon
LOS_ENABLE (Bit D4 in Register 0x74) going low, the peak-to-
peak amplitude in millivolts is loaded into LOS_DATA
(Register 0x36). The contents of LOS_DATA change only when
LOS_ENABLE (Bit D4 in Register 0x74) is toggled low to high
to low while pointing to LOS_ADDRESS[2:0] (Bits[D2:D0] in
Register 0x74) = 7.
PASSIVE EQUALIZER
A passive equalizer (EQ) is available at the input to equalize
large signals that have undergone distortion due to PCB traces,
vias, and connectors. The adaptive equalizer of the ADN2917 is
a factory set default function. If needed, the EQ can be
manually set.
The equalizer can be manually set through the LA_EQ Register
(Register 0x16). An adaptive loop is also available that
optimizes the EQ setting based on characteristics of the
received eye at the phase detector. If the channel is known in
advance, set the EQ setting manually to obtain the best
performance; however, the adaptive EQ finds the best setting in
most cases.
Table 23 indicates a typical EQ setting for several trace lengths.
The values in Table 23 are based on measurements taken on a
test board with simple FR-4 traces. Table 24 lists the typical
maximum reach in inches of FR-4 of the EQ at several data
rates. If a real channel includes lossy connectors or vias, the
FR-4 reach length is shorter. For any real-world system, it is
highly recommended to test several EQ settings with the real
channel to ensure best signal integrity.
Table 23. EQ Settings vs. Trace Length on FR-4
Trace Length (Inches) Typical EQ Setting
6 10
10 12
15 14
20 to 30 15
Table 24. Typical EQ Reach on FR-4 vs. Maximum Data
Rates Supported
Maximum Data Rate
(Gbps)
Typical EQ Reach on FR-4
(Inches)
4 30
8 20
10 15
11
10
Rev. 0 | Page 22 of 32
Data Sheet ADN2917
0 dB EQ
The 0 dB EQ path connects the input signal directly to the
digital logic inside the ADN2917. This is useful at lower data
rates where the signal is large (therefore, the limiting amplifier
is not needed, and power can be saved by deselecting the
limiting amplifier) and unimpaired (therefore, the equalizer is not
needed). The signal swing of the internal digital circuit is
600 mV p-p differential, the minimum signal amplitude that must
be provided as the input in 0 dB EQ mode.
In 0 dB EQ mode, the internal 50 Ω termination resistors can be
configured in one of two ways, either floated or tied to VCC = 1.2 V
(see Figure 26 and Table 28). By setting the RX_TERM_FLOAT
(Bit D7 in Register 0x16) to 1, these 50 Ω termination resistors
are floated internal to the ADN2917 (see Figure 26 and
Figure 29). By setting the RX_TERM_FLOAT bit to 0, these
50 Ω termination resistors are connected to VCC = 1.2 V (see
Figure 26 and Figure 30). In both of these termination cases, the
user must ensure a valid common-mode voltage on the input.
In the case where the termination is floated, the two 50 Ω
resistors are purely a differential termination. The input must
conform to the range of signals shown in Figure 32 and
Figure 33.
In the case of termination to a 1.2 V VCC power supply (see
Figure 30 and Figure 31), the common-mode voltage is created
by joint enterprise between the driver circuit and the 50 Ω
resistors on the ADN2917. For example, the driver can be an
open-drain switched current (see Figure 30), and the 50 Ω
resistors return this current to VCC. In Figure 30, the common-
mode voltage is created by both the current and the resistors. In
this case, ensure that the current is a minimum of 6 mA, which
gives a single-ended swing of 300 mV or a differential swing of
600 mV p-p differential, with VCM = 1.05 V (see Figure 32). The
maximum current is 10 mA, which gives a single-ended
500 mV swing and differential 1.0 V p-p, with VCM = 0.95 V (see
Figure 33).
Another possibility is to have the switched current driver back
terminated, as shown in Figure 31, and the two VCC supplies
having the same potential. In this example, the current is
returned to VCC by two 50 Ω resistors in parallel, or 25 Ω, so
that the minimum current is 12 mA and the maximum current
is 20 mA.
LOCK DETECTOR OPERATION
The lock detector on the ADN2917 has three modes of opera-
tion: normal mode, LTR mode, and static LOL mode.
Normal Mode
In normal mode, the ADN2917 is a continuous rate CDR that
locks onto any data rate from 8.5 Gbps to 11.3 Gbps without the
use of a reference clock as an acquisition aid. In this mode, the
lock detector monitors the frequency difference between the
DCO and the input data frequency, and deasserts the loss of
lock signal, which appears on LOL, Pin 6, when the DCO is
within 250 ppm of the data frequency. This enables the digital
PLL (D/PLL), which pulls the DCO frequency in the remaining
amount and acquires phase lock. When locked, if the input
frequency error exceeds 1000 ppm (0.1%), the loss of lock signal
is reasserted and control returns to the frequency loop, which
begins a new frequency acquisition. The LOL pin remains
asserted until the DCO locks onto a valid input data stream to
within 250 ppm frequency error. This hysteresis is shown in
Figure 22.
11778-022
LOL
0–250 250 1000
f
DCO
ERROR
(ppm)
–1000
1
Figure 22. Transfer Function of LOL
Look to Reference (LTR) Mode
In LTR mode, a reference clock is used as an acquisition aid to
lock the ADN2917 DCO. Lock to reference mode is enabled by
setting CDR_MODE[2:0] (Bits[D6:D4] in Register 0x8) to 3.
The user must also write to FREF_RANGE[1:0] (Bits[D5:D4] in
Register 0xF) and DATA_TO_REF_RATIO[3:0] (Bits[D3:D0]
in Register 0xF) in the LTR_MODE register (Register 0xF) to
set the reference frequency range and the divide ratio of the
data rate with respect to the reference frequency. Finally, the
reference clock power-down to the reference clock buffer must
be deasserted by writing a 0 to I2C the REFCLK_PDN bit
(Bit D2 in Register 0xA). To maintain fastest acquisition,
keep Bit D0 in CTRLC (Register 0xA) set to 1.
For more details, see the Reference Clock (Optional) section.
In this mode, the lock detector monitors the difference in fre-
quency between the divided down DCO and the divided down
reference clock. The loss of lock signal, which appears on LOL
(Pin 6), is deasserted when the DCO is within 250 ppm of the
desired frequency. This enables the D/PLL, which pulls in the
DCO frequency the remaining amount with respect to the input
data and acquires phase lock. When locked, if the frequency
error exceeds 1000 ppm (0.1%), the loss of lock signal is
reasserted and control returns to the frequency loop, which
reacquires with respect to the reference clock. The LOL pin
remains asserted until the DCO frequency is within 250 ppm of
the desired frequency. This hysteresis is shown in Figure 22.
Static LOL Mode
The ADN2917 implements a static LOL feature that indicates if
a loss of lock condition has ever occurred and remains asserted,
even if the ADN2917 regains lock, until the static LOL bit (Bit
D2 in Register 0x6) is manually reset. If there is ever an
occurrence of a loss of lock condition, this bit is internally
asserted to logic high. The static LOL bit remains high even
after the ADN2917 has reacquired lock to a new data rate. This
bit can be reset by writing a 1, followed by 0, to the reset static
LOL bit (Bit D2 in Register 0x8). When reset, the static LOL bit
Rev. 0 | Page 23 of 32
ADN2917 Data Sheet
(Bit D2 in Register 0x6) remains deasserted until another loss of
lock condition occurs.
Writing a 1 to the LOL_CONFIG bit (Bit D4 in Register 0x9)
causes the LOL pin, Pin 6, to become a static LOL indicator. In
this mode, the LOL pin mirrors the contents of the static LOL
bit (Bit D2 in Register 0x6) and has the functionality described
previously. The LOL_CONFIG bit (Bit D4 of Register 0x9)
defaults to 0. In this mode, the LOL pin operates in the normal
operating mode; that is, it is asserted only when the ADN2917
is in acquisition mode and deasserts when the ADN2917 has
reacquired lock.
OUTPUT DISABLE AND SQUELCH
The ADN2917 has two types of output disable/squelch. The
DATOUTP/DATOUTN and CLKOUTP/CLKOUTN outputs
can be disabled by setting DATOUT_DISABLE (Bit D4 in
Register 0x1E) and CLKOUT_DISABLE (Bit D3 in Register 0x1E)
high, respectively. When an output is disabled, it is fully
powered down, saving approximately 30 mW per output.
Disabling DATOUTP/DATOUTN also disables the CLKOUTP/
CLKOUTN output, saving a total of about 60 mW of power.
If it is desired to gate the data output while leaving the clock on,
the output data can be squelched by setting the data squelch bit
(Bit D5 in Register 0x1E) high. In this mode, the data driver is
left powered, but the data itself is forced to be always 0 (or 1),
depending on the setting of the DATA_POLARITY bit (Bit D1
in Register 0x1E).
I2C INTERFACE
The ADN2917 supports a 2-wire, I2C-compatible serial bus,
driving multiple peripherals. Two inputs, serial data (SDA) and
serial clock (SCK), carry information between any devices con-
nected to the bus. Each slave device is recognized by a unique
address. The slave address consists of the seven MSBs of an
8-bit word. The upper six bits (Bits[6:1]) of the 7-bit slave
address are factory programmed to 100000. The LSB of the
slave address (Bit 0) is set by Pin 22, I2C_ADDR. The LSB of the
word sets either a read or write operation (see Figure 13). Logic 1
corresponds to a read operation, whereas Logic 0 corresponds
to a write operation.
To control the device on the bus, the following protocol must be
used. First, the master initiates a data transfer by establishing a
start condition, defined by a high-to-low transition on SDA
while SCK remains high. This indicates that an address/data
stream follows. All peripherals respond to the start condition and
shift the next eight bits (the 7-bit address and the R/W bit).
The bits are transferred from MSB to LSB. The peripheral that
recognizes the transmitted address responds by pulling the
data line low during the ninth clock pulse. This is known as an
acknowledge bit. All other devices withdraw from the bus at
this point and maintain an idle condition. The idle condition is
when the device monitors the SDA and SCK lines waiting for
the start condition and correct transmitted address. The R/W
bit determines the direction of the data. Logic 0 on the LSB of
the first byte means that the master writes information to the
peripheral. Logic 1 on the LSB of the first byte means that the
master reads information from the peripheral.
The ADN2917 acts as a standard slave device on the bus. The
data on the SDA pin is eight bits long, supporting the 7-bit
addresses plus the R/W bit. The ADN2917 has subaddresses to
enable the user-accessible internal registers (see Table 7).
The ADN2917, therefore, interprets the first byte as the device
address and the second byte as the starting subaddress. Auto-
increment mode is supported, allowing data to be read from or
written to the starting subaddress and each subsequent address
without manually addressing the subsequent subaddress. A data
transfer is always terminated by a stop condition. The user can
also access any unique subaddress register on a one-by-one
basis without updating all registers.
Stop and start conditions can be detected at any stage of the
data transfer. If these conditions are asserted out of sequence
with normal read and write operations, they cause an immedi-
ate jump to the idle condition. During a given SCK high period,
issue one start condition, one stop condition, or a single stop
condition followed by a single start condition. If an invalid subad-
dress is issued by the user, the ADN2917 does not issue an
acknowledge and returns to the idle condition. If the user exceeds
the highest subaddress while reading back in auto-increment
mode, the highest subaddress register contents continue to be
output until the master device issues a no acknowledge. This
indicates the end of a read. In a no acknowledge condition, the
SDA line is not pulled low on the ninth pulse. See Figure 14 and
Figure 15 for sample write and read data transfers, respectively,
and Figure 16 for a more detailed timing diagram.
REFERENCE CLOCK (OPTIONAL)
A reference clock is not required to perform clock and data
recovery with the ADN2917. However, support for an optional
reference clock is provided. The reference clock can be driven
differentially or single-ended. If the reference clock is not being
used, float both REFCLKP and REFCLKN.
Two 50 Ω series resistors present a differential load between
REFCLKP and REFCLKN. Common mode is internally set to
0.56 × VCC by a resistor divider between VCC and VEE. See
Figure 23, Figure 24, and Figure 25 for sample configurations.
The reference clock input buffer accepts any differential signal
with a peak-to-peak differential amplitude of greater than
100 mV. Phase noise and duty cycle of the reference clock are
not critical and 100 ppm accuracy is sufficient.
Rev. 0 | Page 24 of 32
Data Sheet ADN2917
50Ω50Ω
VCC/2
REFCLKP
REFCLKN
ADN2917
11778-023
BUFFER
24
23
Figure 23. DC-Coupled, Differential REFCLKx Configuration
50Ω50Ω
VCC/2
VCC
REFCLKP
OUT REFCLKN
ADN2917
11778-024
BUFFER
24
23
CLK
OSC
Figure 24. AC-Coupled, Single-Ended REFCLKx Configuration
50Ω50Ω
VCC/2
REFCLKP
REFCLKN
ADN2917
11778-025
BUFFER
24
23
REFCLK
Figure 25. AC-Coupled, Differential REFCLKx Configuration
The reference clock can be used either as an acquisition aid for
the ADN2917 to lock onto data, or to measure the frequency
of the incoming data to within 0.01%. The modes are mutually
exclusive because, in the first use, the user can force the device
to lock onto only a known data rate; in the second use, the user
can measure an unknown data rate.
Lock to reference mode is enabled by writing a 3 to
CDR_ MODE[2:0] (Bits[6:4] in Register 0x8). An on-chip clock
buffer must be powered on by writing a 0 to the REFCLK_PDN
bit (Bit D2 in Register 0xA). Fine data rate readback mode is
enabled by writing a 1 to the RATE_MEAS_EN bit (Bit D1 in
Register 0x8). Enabling lock to reference and data rate readback
at the same time causes an indeterminate state and is not
supported.
Using the Reference Clock to Lock onto Data
In this mode, the ADN2917 locks onto a frequency derived
from the reference clock according to the following equation:
Data Rate/2(LTR_MODE[3:0] − 1) = REFCLK/2LTR_MODE[5:4]
The user must know exactly what the data rate is and provide a
reference clock that is a function of this rate. The ADN2917 can
still be used as a continuous rate device in this configuration if
the user has the ability to provide a reference clock that has a
variable frequency (see the AN-632 Application Note).
The reference clock can be anywhere between 11.05 MHz and
176.8 MHz. By default, the ADN2917 expects a reference clock
of between 11.05 MHz and 22.1 MHz. If it is between 22.1 MHz
and 44.2 MHz, 44.2 MHz and 88.4 MHz, or 88.4 MHz and
176.8 MHz, the user must configure the ADN2917 to use the
correct reference frequency range by setting the two bits of
FREF_RANGE[1:0] (Bits[D5:D4] in Register 0xF).
Table 25. LTR_MODE (Register 0xF) Settings
LTR_MODE[5:4] Range (MHz) LTR _MODE[3:0] Ratio
00 11.05 to 22.1 0000 2−1
01 22.1 to 44.2 0001 2
0
10 44.2 to 88.4 n 2
n − 1
11 88.4 to 176.8 1010 29
The user can specify a fixed integer multiple of the reference clock
to lock onto using DATA_TO_REF_RATIO[3:0] (Bits[D3:D0]
in Register 0xF). Set
DATA_TO_REF_RATIO[3:0] = data rate ÷ DIV_fREF
where DIV_fREF represents the divided-down reference referred
to the 11.05 MHz to 22.1 MHz band.
For example, if the reference clock frequency is 38.88 MHz and
the input data rate is 9953.28 Mbps, then FREF_RANGE[1:0]
(Bits[D5:D4] in Register 0xF) is set to 01 to give a divided-down
reference clock of 19.44 MHz. DATA_TO_REF_RATIO[3:0]
(Bits[D3D:0]) in Register 0xF) is set to 1010, that is, 10, because
9953.28 Mbps/19.44 MHz = 2(10 − 1)
While the ADN2917 is operating in lock to reference mode, if
the user changes the reference frequency, that is, the fREF range
(Bits[D5:D4] in Register 0xF) or the fREF ratio (Bits[D3:D0] in
Register 0xF), this must be followed by writing a 0-1-0
transition into the INIT_FREQ_ACQ (Bit D6 in Register 0x9)
to initiate a new lock to reference command.
By default in lock to reference clock mode, when lock has been
achieved and the ADN2917 is in tracking mode, the frequency
of the DCO is being compared to the frequency of the reference
clock. If this frequency error exceeds 1000 ppm, lock is lost, LOL is
asserted, and it relocks to the reference clock while continuing
to output a stable clock.
An alternative configuration is enabled by setting the LOL data
bit (Bit D6 of Register 0xF) = 1. In this configuration, when the
device is in tracking mode, the frequency of the DCO is being
compared to the frequency of the input data, rather than the
frequency of the reference clock. If this frequency error exceeds
1000 ppm, lock is lost, LOL is asserted, and it relocks to the
reference clock while continuing to output a stable clock.
Using the Reference Clock to Measure Data Frequency
The user can also provide a reference clock to measure the
recovered data frequency. In this case, the user provides a
reference clock, and the ADN2917 compares the frequency of
the incoming data to the incoming reference clock and returns a
ratio of the two frequencies to 0.01% (100 ppm). The accuracy
error of the reference clock is added to the accuracy of the
ADN2917 data rate measurement. For example, if a 100 ppm
accuracy reference clock is used, the total accuracy of the
measurement is 200 ppm.
Rev. 0 | Page 25 of 32
ADN2917 Data Sheet
The reference clock can range from 11.05 MHz to 176.8 MHz.
Before reading back the data rate using the reference clock,
FREF_RANGE[1:0] (Bits[D5:D4] in Register 0xF) must be set
to the appropriate frequency range with respect to the reference
clock being used according to Table 25. A fine data rate readback
is then executed as follows:
1. Apply the reference clock.
2. Write a 0 to REFCLK_PDN (Bit D2 in Register 0xA) to
enable the reference clock circuit.
3. Write to FREF_RANGE[1:0] (Bits[D5:D4] in Register 0xF)
to select the appropriate reference clock frequency circuit.
4. Write a 1 to RATE_MEAS_EN (Bit D1 in Register 0x8). This
enables the fine data rate measurement capability of the
ADN2917. This bit is level sensitive and does not need to be
reset to perform subsequent frequency measurements.
5. Write a 0-1-0 to RATE_MEAS_RESET (Bit D0 in
Register 0x8). This initiates a new data rate measurement.
6. Read back RATE_MEAS_COMP (Bit D0 in Register 0x6). If
it is 0, the measurement is not complete. If it is 1, the
measurement is complete and the data rate can be read
back on RATE_FREQ[23:0] (Bits[D7:D0] in Registers 0x0,
Register 0x1, and Register 0x2) and FREQ_RB2 (Register 0x5)
(see Table 7). The approximate time for a data rate
measurement is given in Equation 2.
Use the following equation to determine the data rate:
[ ]
( )
DIVRATEFULLRATELTR
REFCLK
DATARATE
f23:0FREQRATE
f2222
_
7]4:5[ ×××
×
=
(1)
where:
fD ATA R ATE is the data rate (Mbps).
RATE_FREQ[23:0] is from FREQ2[7:0] (most significant byte),
FREQ1[7:0], and FREQ0[7:0] (least significant byte). See Table 7.
fREFCLK is the reference clock frequency (MHz).
FULLRATE = FREQ_RB2[6] (Bit D6 in Register 0x5).
DIVRATE = FREQ_RB2[5:2] (Bits[D5:D2] in Register 0x5).
MSB LSB
D23 to D16 D15 to D8 D7 to D0
FREQ2[7:0] FREQ1[7:0] FREQ0[7:0]
Consider an example of a 9.953 Gbps (OC-192) input signal
and a reference clock source of 19.44 MHz at the PIN/NIN and
REFCLKP/ REFCLKN ports, respectively. In this case,
FREF_RANGE[1:0] (Bits[D5:D4] in Register 0xF) = 00, and the
reference frequency falls into the range of 11.05 MHz to
22.1 MHz. After following Step 1 through Step 6, the readback
value of RATE_FREQ[23:0] (Bits[D7:D0] in Register 0x0,
Register 0x1, and Register 0x2) is 0x00FFFD, which is equal to
65533. The readback value of FULLRATE (Bit D6 in Register 0x5)
is 0, and the readback value of DIVRATE[3:0] (Bits[D5:D2] in
Register 0x5) is 0. Entering these values into Equation 1 yields
((65533) × (19.44 × 106))/(20 × 27 × 20 × 20) = 9.95282 Gbps
If subsequent frequency measurements are required, keep
RATE_MEAS_EN (Bit D1 in Register 0x8) set to 1. It does not
need to be reset. The measurement process is reset by writing a
1 followed by a 0 to RATE_MEAS_RESET (Bit D0 in Register 0x8).
This initiates a new data rate measurement. Follow Step 2
through Step 6 to read back the new data rate. Note that a data
rate readback is valid only if the LOL pin is low. If LOL is high,
the data rate readback is invalid.
Initiating a frequency measurement by writing a 0-1-0 to
RATE_MEAS_RESET (Bit D0 in Register 0x8) also resets the
RATE_ MEAS_COMP (Bit D0 in Register 0x6) bit. The
approximate time to complete a frequency measurement from
RATE_MEAS_RESET being written with a 0-1-0 transition to
when the RATE_MEAS_COMP bit returns high is given by
REFCLK
LTR
f
tTimeMeasuremen
]4:5[11 22 ×
=
(2)
LOS Configuration
The LOS detector output, LOS (Pin 5), can be configured to
be either active high or active low. If LOS polarity (Bit D2 in
Register 0x9) is set to Logic 0 (default), the LOS pin is active
high when a loss of signal condition is detected.
ADDITIONAL FEATURES AVAILABLE VIA THE I2C
INTERFACE
Coarse Data Rate Readback
The data rate can be read back over the I2C interface to approx-
imately ±5% without needing an external reference clock
according to the following formula:
DIVRATEFULLRATE
DCO
f
Data 22 ×
=
(3)
where
fDCO is the frequency of the DCO, derived as shown in Table 26.
FULLRATE is from Bit D6 in Register 0x5.
DIVRATE is from Bits[D5:D2] in Register 0x5.
Four oscillator cores defined by VCOSEL[9:8] (Bits[D1:D0] in
Register 0x5) span the highest octave of data rates according to
Table 26.
Table 26. DCO Center Frequency vs. VCOSEL[9:8]
(Bits[D1:D0] in Register 0x5)
Core =
VCOSEL[9:8]
Min Frequency
(MHz) =
Min_f(core)
Max Frequency
(MHz) = Max_f(core)
0 5570 7105
1 7000 8685
2 8610 10,330
3 10,265 11,625
Rev. 0 | Page 26 of 32
Data Sheet ADN2917
Rev. 0 | Page 27 of 32
fDCO is determined from VCOSEL[9:0] (Bits[D7:D0] in Register
0x4 and Bits[D1:D0] in Register 0x5), according to the
following formula:
fDCO =
]VCOSEL[7:0
corefMincorefMax
corefMin
256
)(_)(_
)(_
Worked Example
Read back the contents of FREQ_RB1 (Register 0x4) and
FREQ_RB2 (Register 0x5). For example, with an 10.3125 Gbps
signal presented to the PIN/NIN ports,
VCOSEL[7:0] = 0x11
FREQ_RB2 = 0x03
FULLRATE (Bit D6 in Register 0x5) = 0
DIVRATE (Bits[D5:D2] in Register 0x5) = 0
core (Bits[D1:D0] in Register 0x5) = 3
then
fDCO =
Mbps31.1035517
256
Mbps)1026511625(
Mbps10265
and
Mbps31.10355
22
Mbps31.10355
00
data
f
Initiate Frequency Acquisition
A frequency acquisition can be initiated by writing a 1 followed
by a 0 to INIT_FREQ_ACQ (Bit D6 in Register 0x9). This
initiates a new frequency acquisition while keeping the
ADN2917 in the operating mode that was previously
programmed in Register 0x8 (CTRLA) Register 0x9 (CTRLB),
and Register 0xA (CTRLC).
PRBS Generator/Receiver
The ADN2917 has an integrated PRBS generator and detector
for system testing purposes. The devices are configurable as
either a PRBS detector or a PRBS generator.
The following steps configure the PRBS detector:
1. Set DATA_RECEIVER_ENABLE (Bit D2 in Register 0x3F) to
1 while also setting DATA_RECEIVER_MODE[1:0]
(Bits[D1:D0] in Register 0x3F) according to the desired PRBS
pattern (0: PRBS7; 1: PRBS15; 2: PRBS31). Setting
DATA_RECEIVER_MODE[1:0] to 3 leads to a one-shot
sampling of recovered data into DATA_LOADED[15:0]
(Bits[D7:D0] in Register 0x42 and Register 0x43).
2. Set DATA_RECEIVER_CLEAR (Bit D3 in Register 0x3F) to 1
followed by 0 to clear PRBS_ERROR (Bit D0 in Register 0x41)
and PRBS_ERROR_COUNT (Bits[D7:D0] in Register 0x40).
3. States of PRBS_ERROR and PRBS_ ERROR_COUNT[7:0]
can be frozen by setting DATA_RECEIVER_ENABLE
(Bit D2 in Register 0x3F) to 0.
The following steps configure the PRBS generator:
1. Set DATA_GEN_EN (Bit D2 in Register 0x39) = 1 to
enable the PRBS generator while also setting
DATA_GEN_MODE[1:0] (Bits[D1:D0] in Register 0x39)
for a desired PRBS output pattern (0: PRBS7; 1: PRBS15; 2:
PRBS31). An arbitrary 32-bit pattern stored as
PROG_DATA[31:0] (Bits[D7:D0] in Register 0x3B,
Register 0x3C, Register 0x3D, and Register 0x3E) is
activated by setting DATA_GEN_MODE[1:0] to 3.
2. Strings of consecutive identical digits of sensed
DATA_CID_BIT (Bit D5 in Register 0x39) can be
introduced in the generator with DATA_CID_EN (Bit D4
in Register 0x39) set to 1. The length of consecutive
identical digits (CIDs) is 8 × DATA_CID_LENGTH[7:0]
(Bits[D7:D0] in Register 0x3A), which is set via
PRBS Gen 2[7:0] register (Register 0x3A).
Table 27. PRBS Settings
PRBS
Patterns
DATA_GEN_MODE[1:0]
(Bits[D1:D0] in Register 0x39)
PRBS
Polynomial
PRBS7 0x00 1 + X6 + X7
PRBS15 0x01 1 + X14 + X15
PRBS31 0x10 1 + X28 + X31
PROG_DATA
[31:0]1
0x11 Not applicable
1 Bits[D7:D0] in Register 0x3B, Register 0x3C, Register 0x3D, and Register 0x3E.
Double Data Rate Mode
The recovered output clock is a double data rate (DDR) clock,
where the output clock frequency is ½ the data rate.
This allows direct interfacing to FPGAs that support clocking
on both rising and falling edges.
Disable Output Buffers
The ADN2917 provides the option of disabling the output buffers
for power savings. The clock output buffer can be disabled by
setting Bit CLKOUT_DISABLE (Bit D3 in Register 0x1E) = 1.
This reduces the total output power by 30 mW. For a total of 60
mW of power savings, such as in a low power standby mode, both
the CLKOUTx and DATOUTx buffers can be disabled together
by setting the DATOUT_DISABLE bit (Bit D4 of 0x1E) = 1.
Transmission Lines
Use of 50 Ω transmission lines is required for all high frequency
input and output signals to minimize reflections: PIN, NIN,
CLKOUTP, CLKOUTN, DATOUTP, and DATOUTN (also
REFCLKP and REFCLKN, if using a high frequency reference
clock, such as 155 MHz). It is also necessary for the PIN and NIN
input traces to be matched in length, and the CLKOUTP,
CLKOUTN, DATOUTP, and DATOUTN output traces to be
matched in length to avoid skew between the differential traces.
The high speed inputs (PIN and NIN) are each internally termi-
nated with 50 Ω to an internal reference voltage (see Figure 26).
As with any high speed, mixed-signal circuit, take care to keep
all high speed digital traces away from sensitive analog nodes.
ADN2917 Data Sheet
The high speed outputs (DATOUTP, DATOUTN, CLKOUTP, and
CLKOUTN) are internally terminated with 50 Ω to VCC.
Soldering Guidelines for Lead Frame Chip Scale Package
The lands on the 24-lead LFCSP are rectangular. The PCB pad
for these lands is 0.1 mm longer than the package land length,
and 0.05 mm wider than the package land width. Center the
land on the pad to ensure that the solder joint size is maximized.
The bottom of the lead frame chip scale package has a central
exposed pad. The pad on the PCB must be at least as large as
this exposed pad. The user must connect the exposed pad to
VEE using plugged vias to prevent solder from leaking through
the vias during reflow. This ensures a solid connection from the
exposed pad to VEE.
It is highly recommended to include as many vias as possible
when connecting the exposed pad to VEE. This minimizes the
thermal resistance between the die and VEE, and minimizes the
die temperature. It is recommended that the vias be connected
to a VEE plane, or planes, rather than a signal trace, to improve
heat dissipation as shown in Figure 27.
Placing an external VEE plane on the backside of the board
opposite the ADN2917 provides an additional benefit because
this allows easier heat dissipation into the ambient environment.
INPUT CONFIGURATIONS
The ADN2917 input stage can work with the signal source in
either ac-coupled or dc-coupled configuration. To best fit in a
required applications environment, the ADN2917 supports one
of following input modes: limiting amplifier, equalizer, or
bypass. It is easy to set the ADN2917 to use any required input
configuration through the I2C bus. Figure 26 shows a block
diagram of the input stage circuit.
A correct input signal pass is configurable with the INPUT_
SEL[1:0] bits (Bits[D6:D5] in Register 0x16). Table 28 shows the
INPUT_SEL[1:0] bits and the input signal configuration.
LOS
DETECT
LA
BYPASS
VCC
FLOAT
VREF
INPUT_SEL[1:0]
RX_TERM_FLOAT
EQ
LOS
PIN
NIN
2
50Ω50Ω2.9kΩ2.9kΩ
11778-026
SLICE
ADJUST
Figure 26. Input Stage Circuit Block Diagram
Table 28. Input Signal Configuration
Selected Input
INPUT_SEL[1:0] (Bits[D6:D5] in
Register 0x16)
RX_TERM_FLOAT (Bit D7 in
Register 0x16) = 0 RX_TERM_FLOAT = 1
Limiting Amplifier 00 V
REF
Not defined
Equalizer 01 V
REF
Not defined
Bypass (0 dB Buffer) 10 V
CC
Float
Not Defined 11 Not defined Not defined
Rev. 0 | Page 28 of 32
Data Sheet ADN2917
PACKAGE
COPPER PLANE—V
EE
PCB PAD
WITH
25 VIAS
HEAT
DISSIPATION
HEAT
DISSIPATION
COPPER PLANE—V
EE
EXPOSED PAD
11778-027
Figure 27. Connecting Vias to VEE
Choosing AC Coupling Capacitors
AC coupling capacitors at the inputs (PIN, NIN) and outputs
(DATOUTP, DATOUTN) of the ADN2917 must be chosen
such that the device works properly over the full range of data
rates used in the application. When choosing the capacitors, the
time constant formed with the two 50 Ω resistors in the signal
path must be considered. When a large number of CIDs are
applied, the capacitor voltage can droop due to baseline wander
(see Figure 28), causing pattern dependent jitter (PDJ).
The user must determine how much droop is tolerable and choose
an ac coupling capacitor based on that amount of droop. The
amount of PDJ can then be approximated based on the capaci-
tor selection. The actual capacitor value selection may require
some trade-offs between droop and PDJ.
For example, assuming that 2% droop is tolerable, the
maximum differential droop is 4%.
Normalizing to V p-p,
Droop = Δ V = 0.04 V = 0.5 V p-p (1 − e–t/τ)
τ = 12t
where:
τ is the RC time constant (C is the ac coupling capacitor, R = 100 Ω
seen by C).
t is the total discharge time
t = nΤ
where:
n is the number of CIDs.
T is the bit period.
Calculate the capacitor value by combining the equations for τ
and t.
C = 12nT/R
When the capacitor value is selected, the PDJ can be
approximated as
PDJps p-p = 0.5tr(1 − e(−nT/RC)/0.6
where:
PDJps p-p is the amount of pattern dependent jitter allowed,
<0.01 UI p-p typical.
tr is the rise time, which is equal to 0.22/BW; BW ≈ 0.7 (bit
rate).
Note that this expression for tr is accurate only for the inputs.
The output rise time for the ADN2917 is ~30 ps regardless of
data rate.
Rev. 0 | Page 29 of 32
ADN2917 Data Sheet
NOTES
1. DURI NG T HE DATA PATTE RNS WI TH HIGH T RANS ITIO N DE NS ITY , DI FFE RE NTI AL DC VO LT AGE AT V1 AND V2 IS ZERO .
2. WHEN T HE TI A OUTP UTS CONSECUTI V E IDENT ICAL DIGIT S , V1 AND V1b ARE DRIVEN TO DIF FERENT DC L E V E LS. V 2 AND V 2b DI S CHARGE T O
THE V
REF
LEVEL, WHICH EFFECTIVELY INTRODUCES A DIFFERENTIAL DC OFFSET ACROSS THE AC COUPLING CAPACITORS.
3. WHEN T HE BURS T O F DATA S TART S AGAIN, T HE DIFFERE NTIAL DC O FF S ET ACRO S S THE AC COUPL ING CAP ACITORS IS APP LI E D TO THE
INPUT LEVELS, CAUSING A DC SHIFT IN THE DIFFERENTIAL INPUT. THIS SHIFT IS LARGE ENOUGH SUCH THAT ONE OF THE STATES, EITHER
HIG H OR L OW , DEPE NDING ON THE LEV E LS O F V1 AND V 1b WHE N THE TIA BE GAN DETECT ING AND OUT P UTTING A CID DATA S Y S TEM , IS
CANCELLED OUT. THE QUANTIZER DOES NOT RECOGNIZE THIS AS A VALID STATE.
4. THE DC OFFSET SLOWLY DISCHARGES UNTIL THE DIFFERENTIAL INPUT VOLTAGE EXCEEDS THE SENSITIVITY OF THE ADN2917. THE
QUANTIZER RECOGNIZES BOTH HIGH AND LOW STATES AT THIS POINT.
V1
V1b
V2
V2b
VDIFF
VDIFF = V 2 – V 2b
VT H = ADN2917 QUANT IZ E R THRESHOL D
2 3 41
VREF
VTH
CDR
V
REF
50Ω
50Ω
PIN
NIN
ADN2917
C
OUT
DATOUTP
DATOUTN
C
IN
V2
V2b
V1
V1b
TIA
VCC
11778-028
2
Figure 28. Example of Baseline Wander
Rev. 0 | Page 30 of 32
Data Sheet ADN2917
Rev. 0 | Page 31 of 32
DC-COUPLED APPLICATION
The inputs to the ADN2917 can also be dc-coupled. This can be
necessary in burst mode applications with long periods of CIDs
and where baseline wander cannot be tolerated. If the inputs to
the ADN2917 are dc-coupled, care must be taken not to violate
the input range and common-mode level requirements of the
ADN2917 (see Figure 32 or Figure 33). If dc coupling is required,
and the output levels of the transimpedance amplifier (TIA) do
not adhere to the levels shown in Figure 32 or Figure 33, level
shifting and/or attenuation must occur between the TIA outputs
and the ADN2917 inputs.
50Ω
50Ω
50Ω
ADN2917
VDD
TIA
PIN
NIN
TIA
11778-029
Figure 29. DC-Coupled Application, 0 dB EQ Input (Rx Term Float Mode)
Figure 30 shows the default dc-coupled situation when using
the 0 dB EQ input. The two terms are connected directly to
VCC in a normal CML fashion, giving a common mode that is
set by the dc signal strength from the driving chip. The 0 dB EQ
input has a high common-mode range and can tolerate VCM up
to and including VCC.
50Ω
50Ω
50Ω
VCC
ADN2917
PIN
I
NIN
11778-030
Figure 30. DC-Coupled Application, 0 dB EQ Input (Normal Mode)
50Ω
50Ω
50Ω
50Ω50Ω
VCC
ADN2917
PIN
V
C
C
I
NIN
11778-031
Figure 31. DC-Coupled Application, 0 dB EQ Input (Back Terminated Mode)
1.2V
0.9V
600mV p-p,
DIFF
600mV p-p,
DIFF
V
CM
= 1.05V
V
CM
= 0.65V
INPUT (V)
0.8V
0.5V
11778-032
Figure 32. Minimum Allowed DC-Coupled Input Levels
1.2V
0.7V
1.0V p-p,
DIFF
1.0V p-p,
DIFF
V
CM
= 0.95V
V
CM
= 0.75V
INPUT (V)
1.0V
0.5V
11778-033
Figure 33. Maximum Allowed DC-Coupled Input Levels
ADN2917 Data Sheet
Rev. 0 | Page 32 of 32
OUTLINE DIMENSIONS
COMPLIANT
TO
JEDEC S TANDARDS M O-220-VG GD-8
04-11-2012-A
1
0.50
BSC
PIN 1
INDICATOR
2.50 RE F
0.50
0.40
0.30
TOP VIEW
12° M AX 0.80 MAX
0.65 TY P
SEATING
PLANE COPLANARITY
0.08
1.00
0.85
0.80
0.30
0.23
0.18
0.05 M AX
0.02 NOM
0.20 RE F
0.25 MI N
2.65
2.50 S Q
2.35
24
7
19
12
13
18
6
0.60 MAX
0.60 M AX
PIN 1
INDICATOR
4.10
4.00 SQ
3.90
3.75 BSC
SQ
EXPOSED
PAD
FOR PROPER CONNECTION OF
THE EXPOSED PAD, REFER TO
THE P IN CONFIG URATI ON AND
FUNCTI ON DES CRIPTI ONS
SECTION OF T HIS DATA SHEET.
BOTTOM VI EW
Figure 34. 24-Lead Lead Frame Chip Scale Package [LFCSP_VQ]
4 mm × 4 mm Body, Very Thin Quad
(CP-24-3)
Dimensions shown in millimeters
ORDERING GUIDE
Model1 Temperature Range Package Description Package Option Ordering Quantity
ADN2917ACPZ −40°C to +85°C 24-Lead LFCSP_VQ CP-24-3 490
ADN2917ACPZ-RL7 −40°C to +85°C 24-Lead LFCSP_VQ CP-24-3 1500
EVALZ-ADN2917
1 Z = RoHS Compliant Part.
I2C refers to a communications protocol originally developed by Philips Semiconductors (now NXP Semiconductors).
©2014 Analog Devices, Inc. All rights reserved. Trademarks and
registered trademarks are the property of their respective owners.
D11778-0-5/14(0)
