Continuous Rate 8.5 Gbps to 11.3 Gbps Clock and Data Recovery IC with Integrated Limiting Amp/EQ ADN2917 Data Sheet FEATURES GENERAL DESCRIPTION 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 x 4 mm 24-lead LFCSP 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 automatically 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 userprogrammable 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 x 4 mm, 24-lead frame chip scale package (LFCSP). All ADN2917 specifications are defined over the ambient temperature range of -40C to +85C, unless otherwise noted. APPLICATIONS SONET/SDH OC-192, 10GFC, and 10GE and all associated FECs XFP, line cards, clocks, routers, repeaters, instruments Any rate regenerators/repeaters FUNCTIONAL BLOCK DIAGRAM SDA LOL I2C REGISTERS FREQUENCY ACQUISITION AND LOCK DETECTOR LOS THRESH SLICE ADJUST I2C_ADDR PIN 2 BYPASS NIN 50 50 I2C VCM CML DDR FIFO /N DOWNSAMPLER AND LOOP FILTER LA DATA INPUT SAMPLER CML CLK SAMPLE PHASE ADJUST LOS DETECT CLKOUTP/ CLKOUTN DATA RATE ADN2917 LOS DATOUTP/ DATOUTN TXD SCK REFCLKP/ REFCLKN (OPTIONAL) /2 DCO RXD RXCK EQ CLOCK I2C PHASE SHIFTER 11778-001 VCC FLOAT 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 (c)2014 Analog Devices, Inc. All rights reserved. Technical Support www.analog.com ADN2917 Data Sheet TABLE OF CONTENTS Features .............................................................................................. 1 Frequency Acquisition ............................................................... 20 Applications ....................................................................................... 1 Limiting Amplifier ..................................................................... 20 General Description ......................................................................... 1 Slice Adjust .................................................................................. 20 Functional Block Diagram .............................................................. 1 Edge Select................................................................................... 20 Revision History ............................................................................... 2 Loss of Signal Detector .............................................................. 22 Specifications..................................................................................... 3 Passive Equalizer ........................................................................ 22 Jitter Specifications ....................................................................... 4 0 dB EQ ........................................................................................ 23 Output and Timing Specifications ............................................. 5 Lock Detector Operation .......................................................... 23 Timing Diagrams.......................................................................... 7 Output Disable and Squelch ..................................................... 24 Absolute Maximum Ratings ............................................................ 8 I2C Interface ................................................................................ 24 Thermal Characteristics .............................................................. 8 Reference Clock (Optional) ...................................................... 24 ESD Caution .................................................................................. 8 Additional Features Available via the I2C Interface ............... 26 Pin Configuration and Function Descriptions ............................. 9 Input Configurations ................................................................. 28 Typical Performance Characteristics ........................................... 10 DC-Coupled Application .......................................................... 31 2 I C Interface Timing and Internal Register Descriptions ......... 12 Outline Dimensions ....................................................................... 32 Register Map ............................................................................... 13 Ordering Guide .......................................................................... 32 Theory of Operation ...................................................................... 18 Functional Description .................................................................. 20 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 DATA RATE SUPPORT RANGE INPUT--DC CHARACTERISTICS Peak-to-Peak Differential Input 1 Input Resistance BYPASS PATH--CML INPUT Input Voltage Range Input Common-Mode Level Differential Input Sensitivity OC-192 8GFC 2 LIMITING AMPLIFIER INPUT PATH Differential Input Sensitivity OC-192 8GFC2 10.3125 Gbps EQUALIZER INPUT PATH Differential Input Sensitivity 8GFC2 OC-192 INPUT--AC CHARACTERISTICS S11 LOS DETECT Loss of Signal Detect Test Conditions/Comments PIN - NIN Differential At PIN or NIN, dc-coupled, RX_TERM_FLOAT = 1 (float) DC-coupled (see Figure 32), 600 mV p-p differential, RX_TERM_FLOAT = 1 (float) DCO Frequency Error for LOL Deassert LOL Assert Response Time ACQUISITION TIME Lock to Data (LTD) Mode Optional LTR Mode 4 DATA RATE READBACK ACCURACY Coarse Readback Fine Readback Typ 95 100 0.5 0.65 Max 11.3 Unit Gbps 1.0 105 V VCC VCC - 0.15 V V AC-coupled, RX_TERM_FLOAT = 0 (VCM = 1.2 V), bit error rate (BER) = 1 x 10-10 Jitter tolerance scrambled pattern (JTSPAT), ac-coupled, RX_TERM_FLOAT = 0 (VCM = 1.2 V), BER = 1 x 10-12 200 mV p-p 200 mV p-p BER = 1 x 10-10 JTSPAT, BER = 1 x 10-12 JTSPAT, BER = 1 x 10-12 9.2 8.3 11.0 mV p-p mV p-p mV p-p 15-inch FR-4, 100 differential transmission line, adaptive EQ on JTSPAT, BER = 1 x 10-12 BER = 1 x 10-10 115 184 mV p-p mV p-p At 7.5 GHz, differential return loss, see Figure 9 -12 dB 10 5 128 5.7 135 110 mV p-p mV p-p mV p-p dB s s 1000 ppm 250 25 18 ppm s s 0.5 0.5 0.5 6.0 ms ms ms ms 5 100 % ppm Loss of signal minimum program value Loss of signal maximum program value Hysteresis (Electrical) LOS Assert Time LOS Deassert Time LOSS OF LOCK (LOL) DETECT DCO Frequency Error for LOL Assert Min 8.5 AC-coupled 3 AC-coupled3 With respect to nominal, data collected in lock to reference (LTR) mode With respect to nominal, data collected in LTR mode 8.5 Gbps, JTSPAT 10 Gbps OC192 11.3 Gbps 8.5 Gbps, JTSPAT In addition to reference clock accuracy Rev. 0 | Page 3 of 32 ADN2917 Parameter POWER SUPPLY VOLTAGE VCC VDD VCC1 POWER SUPPLY CURRENT VCC VDD VCC1 TOTAL POWER DISSIPATION Clock Output Enabled Clock Output Disabled Data Sheet Test Conditions/Comments Min Typ Max Unit 1.14 2.97 1.62 1.2 3.3 1.8 1.26 3.63 3.63 V V V Limiting amplifier mode, clock output enabled 8GFC,2 JTSPAT OC-192 8GFC,2 JTSPAT OC-192 8GFC,2 JTSPAT OC-192 319.1 333 7.20 7.21 22.2 35.1 359.5 377.4 8.1 8.59 28.4 47.4 mA mA mA mA mA mA Limiting amplifier mode, 8.5 Gbps Limiting amplifier mode, 9.953 Gbps Equalizer mode, 8.5 Gbps Equalizer mode, 11.3 Gbps 446.6 486.5 352 430 +85 mW mW mW mW C OPERATING TEMPERATURE RANGE -40 1 See Figure 33. Fibre Channel Physical Interface 4 standard, FC-PI-4, Rev 8.00, May 21, 2008. When ac-coupled, the LOS assert and deassert times are dominated by the RC time constant of the ac coupling capacitor and the 100 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. 2 3 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 differential termination load, I2C register default settings, unless otherwise noted. Table 2. Parameter PHASE-LOCKED LOOP CHARACTERISTICS Jitter Transfer Bandwidth (BW) 1 OC-192 8GFC 3 Jitter Peaking OC-192 8GFC3 Jitter Generation OC-192 8GFC3 Jitter Tolerance OC-192 Test Conditions/Comments Typ Max Unit TRANBW[2:0] = 3 OTN mode, 2 TRANBW[2:0] = 1 1064 294 1242 1650 529 1676 kHz kHz kHz 20 kHz to 80 MHz 20 kHz to 80 MHz 0.014 0.004 0.024 0.021 dB dB Unfiltered Unfiltered Unfiltered Unfiltered TRANBW[2:0] = 4 (default) 2000 Hz 20 kHz 400 kHz 4 MHz 80 MHz 0.0045 0.076 0.005 0.044 0.0067 UI rms UI p-p UI rms UI p-p Rev. 0 | Page 4 of 32 Min 0.36 0.28 4255 106 3.78 0.50 0.43 UI p-p UI p-p UI p-p UI p-p UI p-p Data Sheet Parameter 8GFC,3 JTSPAT Sinusoidal Jitter at 340 kHz Sinusoidal Jitter at 5.098 MHz Sinusoidal Jitter at 80 MHz Rx Jitter Tracking Test 4 ADN2917 Test Conditions/Comments Min Typ Max Unit 6.7 0.53 0.59 UI p-p UI p-p UI p-p <10-12 <10-12 BER BER 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 frequency 5 10-12 10-12 510 kHz, 1 UI 100 kHz, 5 UI 1 Jitter transfer bandwidth is programmable by adjusting TRANBW[2:0] in the DPLLA register (Register 0x10). 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. 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. 2 3 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 differential termination load, I2C register default settings, unless otherwise noted. Table 3. Parameter CML OUTPUT CHARACTERISTICS Data Differential Output Swing Test Conditions/Comments Min Typ Max Unit 535 600 672 mV p-p 668 189 406 724 219 508 771 252 570 mV p-p mV p-p mV p-p Output High Voltage OC-192, DATA_SWING[3:0] (Bits[D7:D4] in Register 0x1F) setting = 0xC (default) OC-192, DATA_SWING[3:0] setting = 0xF (maximum) OC-192, DATA_SWING[3:0] setting = 0x4 (minimum) OC-192, CLOCK_SWING[3:0] (Bits[D3:D0] in Register 0x1F) setting = 0xC (default) OC-192, CLOCK_SWING[3:0] setting = 0xF (maximum) OC-192, CLOCK_SWING[3:0] setting = 0x4 (minimum) 8GFC, DATA_SWING[3:0] setting = 0xC (default) 8GFC, DATA_SWING[3:0] setting = 0xF (maximum) 8GFC, DATA_SWING[3:0] setting = 0x4 (minimum 8GFC, CLOCK_SWING[3:0] setting = 0xC (default) 8GFC, CLOCK_SWING[3:0] setting = 0xF (maximum) 8GFC, CLOCK_SWING[3:0] setting = 0x4 (minimum) VOH, dc-coupled 448 162 540 662 190 426 489 166 VCC - 0.05 659 249 666 778 245 588 680 245 VCC mV p-p mV p-p mV p-p mV p-p mV p-p mV p-p mV p-p mV p-p V Output Low Voltage VOL, dc-coupled VCC - 0.36 583 217 600 725 214 518 603 213 VCC - 0.025 VCC - 0.325 VCC - 0.29 V 20% to 80%, at OC-192, DATOUTN/DATOUTP 20% to 80%, at OC-192, CLKOUTN/CLKOUTP 20% to 80%, at 8GFC, 1 DATOUTN/DATOUTP 20% to 80%, at 8GFC,1 CLKOUTN/CLKOUTP 80% to 20%, at OC-192, DATOUTN/DATOUTP 80% to 20%, at OC-192, CLKOUTN/CLKOUTP 80% to 20%, at 8GFC,1 DATOUTN/DATOUTP 80% to 20%, at 8GFC,1 CLKOUTN/CLKOUTP tS (see Figure 2) tH (see Figure 2) tS (see Figure 3) tH (see Figure 3) 17.4 22.2 20.4 23.1 17.5 23.9 23 25 46.5 33.1 44 35.8 49.1 33.7 46.8 37.1 ps ps ps ps ps ps ps ps UI UI UI UI Clock Differential Output Swing Data Differential Output Swing Clock Differential Output Swing CML OUTPUT TIMING CHARACTERISTICS Rise Time Fall Time Setup Time, Full Rate Clock Hold Time, Full Rate Clock Setup Time, DDR Clock Hold Time, DDR clock Rev. 0 | Page 5 of 32 32.6 28.3 33.1 29.7 33 29.2 34.2 31.3 0.5 0.5 0.5 0.5 ADN2917 Parameter I2C INTERFACE DC CHARACTERISTICS Input High Voltage Input Low Voltage Input Current Output Low Voltage I2C INTERFACE TIMING SCK Clock Frequency SCK Pulse Width High SCK Pulse Width Low Start Condition Hold Time Start Condition Setup Time Data Setup Time Data Hold Time SCK/SDA Rise/Fall Time 2 Stop Condition Setup Time Bus Free Time Between Stop and Start Conditions LVTTL DC INPUT CHARACTERISITICS (I2C_ADDR) Input Voltage High Low Input Current High Low LVTTL DC OUTPUT CHARACTERISITICS (LOS/LOL) Output Voltage High Low REFERENCE CLOCK CHARACTERISTICS Input Compliance Voltage (SingleEnded) Minimum Input Drive Reference Frequency Required Accuracy 3 Data Sheet Test Conditions/Comments LVTTL VIH VIL VIN = 0.1 x VDD or VIN = 0.9 x VDD VOL, IOL = 3.0 mA See Figure 17 Min tHIGH tLOW tHD;STA tSU;STA tSU;DAT tHD;DAT tR/tF tSU;STO tBUF 600 1300 600 600 100 300 20 + 0.1 Cb 600 1300 VIH VIL 2.0 Typ Max Unit 0.8 +10.0 0.4 V V A V 2.0 -10.0 400 300 0.8 IIH, VIN = 2.4 V IIL, VIN = 0.4 V +5 -5 VOH, IOH = 2.0 mA VOL, IOL = -2.0 mA Optional LTR mode VCM (no input offset, no input current), see Figure 25, ac-coupled input See Figure 25, ac-coupled, differential input 2.4 0.55 AC-coupled, differential input 1 100 V V A A 0.4 V V 1.0 V 176.8 mV p-p diff MHz ppm 100 11.05 kHz ns ns ns ns ns ns ns ns ns Fibre Channel Physical Interface 4 standard, FC-PI-4, Rev 8.00, May 21, 2008. 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. 2 Rev. 0 | Page 6 of 32 Data Sheet ADN2917 TIMING DIAGRAMS CLKOUTP tH 11778-002 tS DATOUTP/ DATOUTN Figure 2. Data to Clock Timing (Full Rate Clock Mode) CLKOUTP tH 11778-003 tS DATOUTP/ DATOUTN Figure 3. Data to Clock Timing (Half Rate Clock/DDR Mode) DATOUTP LOGIC HIGH LOGIC LOW LOGIC HIGH LOGIC LOW LOGIC HIGH LOGIC LOW VSE VSE 0V DATOUTP - DATOUTN LOGIC HIGH VDIFF LOGIC LOW Figure 4. Single-Ended vs. Differential Output Amplitude Relationship Rev. 0 | Page 7 of 32 11778-004 DATOUTN ADN2917 Data Sheet ABSOLUTE MAXIMUM RATINGS THERMAL CHARACTERISTICS Table 4. Parameter Supply Voltage (VCC = 1.2 V) Supply Voltage (VDD and VCC1 = 3.3 V) Maximum Input Voltage (REFCLKP/REFCLKN, NIN/PIN) Minimum Input Voltage (REFCLKP/REFCLKN, NIN/PIN) Maximum Input Voltage (SDA, SCK, I2C_ADDR) Minimum Input Voltage (SDA, SCK, I2C_ADDR) Maximum Junction Temperature Storage Temperature Range Lead Temperature (Soldering, 10 sec) Thermal Resistance Rating 1.26 V 3.63 V 1.26 V 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. VEE - 0.4 V Table 5. Thermal Resistance 3.63 V Package Type 24-Lead LFCSP_VQ VEE - 0.4 V 1 125C -65C to +150C 300C 2 3 Junction to ambient. Junction to base. Junction to case. ESD CAUTION 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. Rev. 0 | Page 8 of 32 JA1 45 JB2 5 JC3 11 Unit C/W Data Sheet ADN2917 24 23 22 21 20 19 REFCLKP REFCLKN I2C_ADDR VCC SDA SCK PIN CONFIGURATION AND FUNCTION DESCRIPTIONS 1 2 3 4 5 6 PIN 1 INDICATOR ADN2917 TOP VIEW (Not to Scale) 18 17 16 15 14 13 VCC VDD DNC DATOUTP DATOUTN VCC NOTES 1. DNC = DO NOT CONNECT. 2. THE EXPOSED PAD ON THE BOTTOM OF THE DEVICE PACKAGE MUST BE CONNECTED TO VEE ELECTRICALLY. THE EXPOSED PAD WORKS AS A HEAT SINK. 11778-005 VEE VCC1 VDD CLKOUTN CLKOUTP VEE 7 8 9 10 11 12 VCC PIN NIN VEE LOS LOL Figure 5. Pin Configuration Table 6. Pin Function Descriptions Pin No. 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 Mnemonic VCC PIN NIN VEE LOS LOL VEE VCC1 VDD CLKOUTN CLKOUTP VEE VCC DATOUTN DATOUTP DNC VDD VCC SCK SDA VCC I2C_ADDR Type 1 P AI AI P DO DO P P P DO DO P P DO DO DI P P DI DIO P DI 23 24 REFCLKN REFCLKP EPAD DI DI P 1 Description 1.2 V Supply for Limiting Amplifier. Positive Differential Data Input (CML). Negative Differential Data Input (CML). Ground for Limiting Amplifier. Loss of Signal Output (Active High). Loss of Lock Output (Active High). Digital Control Oscillator (DCO) Ground. 1.8 V to 3.3 V DCO Supply. 3.3 V High Supply. Negative Differential Recovered Clock Output (CML). Positive Differential Recovered Clock Output (CML). Ground for CML Output Drivers. 1.2 V Supply for CML Output Drivers. Negative Differential Retimed Data Output (CML). Positive Differential Retimed Data Output (CML). Do Not Connect. Tie off to ground. Leave this pin floating. 3.3 V High Supply. 1.2 V Core Digital Supply. Clock for I2C. Bidirectional Data for I2C. 1.2 V Core Supply. I2C Address. Sets the device I2C address = 0x80 when I2C_ADDR = 0, and the device I2C address = 0x82 when I2C_ADDR = 1. Negative Reference Clock Input (Optional). Positive Reference Clock Input (Optional). 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. 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 = 25C, 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. 1k ADN2917 TOLERANCE 61.4mV/DIV JITTER AMPLITUDE (UI) 100 10 1 SONET REQUIREMENT MASK 0.01 100 16.8ps/DIV 1k 10k 100k 1M 10M 100M JITTER FREQUENCY (Hz) Figure 6. Output Eye Diagram at OC-192 11778-008 11778-006 0.1 Figure 8. Jitter Tolerance: OC-192 5 0 XFP MASK 0 -5 LOG MAGNITUDE (dB) -10 -10 -15 -20 -25 -30 -15 -20 -25 -30 -35 -45 1k 10k 100k 1M 10M FREQUENCY (Hz) 100M -40 1M 10M 100M 1G 10G FREQUENCY (Hz) Figure 9. Typical S11 Spectrum Performance Figure 7. Jitter Transfer: OC-192, TRANBW[2:0] (Bits[D2:D0] in Register 0x10) = 3 Rev. 0 | Page 10 of 32 100G 11778-009 -35 -40 11778-007 JITTER TRANSFER (dB) -5 Data Sheet ADN2917 0.6 16 14 SENSITIVITY (mV p-p DIFF) 0.5 BER 0.4 0.3 0.2 TYPICAL ADAPTIVE EQ SETTING 12 10 8 6 4 0.1 8 10 12 14 16 EQ SETTING 0 DATA RATE (Gbps) 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 8 6 4 2 DATA RATE (Gbps) 11778-010 1.07090 x 1010 9.95328 x 109 0 2.66600 x 109 SENSITIVITY (mV p-p DIFF) 10 Figure 11. Sensitivities of SONET/SDH Data Rates (BER = 10-10) Rev. 0 | Page 11 of 32 11778-012 6 1.10957 x 1010 4 1.03125 x 1010 2 9.95328 x 109 0 8.50000 x 109 0 11778-011 2 Figure 12. Sensitivities of Non-SONET/SDH Data Rates (BER = 10-12) ADN2917 Data Sheet I2C INTERFACE TIMING AND INTERNAL REGISTER DESCRIPTIONS R/W CTRL. SLAVE ADDRESS[6:0] 0 0 0 0 0 x x 11778-013 1 MSB = 1 SET BY 0 = W PIN 22 1 = R S SLAVE ADDR, LSB = 0 (W) DATA A(S) A(S) SUBADDR A(S) DATA A(S) P 11778-014 Figure 13. Slave Address Configuration 2 Figure 14. I C Write Data Transfer SLAVE ADDR, LSB = 0 (W) A(S) SUBADDR A(S) S SLAVE ADDR, LSB = 1 (R) A(S) DATA A(M) DATA A(M) P = STOP BIT A(M) = NO ACKNOWLEDGE BY MASTER A(M) = ACKNOWLEDGE BY MASTER P 11778-015 S S = START BIT A(S) = ACKNOWLEDGE BY SLAVE Figure 15. I2C Read Data Transfer SDA A6 A5 STOP BIT DATA SUBADDRESS SLAVE ADDRESS A7 A0 D7 D0 SCK S WR ACK ACK ACK DATA[6:1] SUBADDR[6:1] SLAVE ADDR[4:0] Figure 16. I2C Data Transfer Timing Diagram tF tSU;DAT tHD;STA tBUF SDA tR tR tSU;STO tF tLOW tHIGH tHD;STA S tSU;STA S tHD;DAT 2 Figure 17. I C Interface Timing Diagram Rev. 0 | Page 12 of 32 P S 11778-017 SCK P 11778-016 START BIT 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 Addr (Hex) Default (Hex) 0x0 0x1 0x2 0x4 0x5 0x6 N/A N/A N/A N/A N/A N/A General Control CTRLA R/W 0x8 0x10 0 CTRLB R/W 0x9 0x00 SOFTWARE_ RESET CTRLC R/W 0xA 0x04 0 INIT_ FREQ_ ACQ 0 FLL Control LTR_MODE R/W 0xF 0x00 0 LOL data DPLLA DPLLD R/W R/W 0x10 0x13 0x1C 0x06 0 0 0 0 Phase Slice R/W W 0x14 0x15 0x00 N/A 0 LA_EQ R/W 0x16 0x08 0 Extended slice RX_ TERM_ FLOAT Slice R Readback Output Control OUTPUTA R/W 0x73 N/A 0x1E 0x00 OUTPUTB R/W 0x1F 0xCC LOS_DATA LOS_THRESH LOS_CTRL R/W R/W R/W 0x36 0x38 0x74 0x00 0x0A 0x00 0 PRBS Control PRBS Gen 1 R/W 0x39 0x00 0 0 DATA_ CID_ BIT PRBS Gen 2 PRBS Gen 3 PRBS Gen 4 PRBS Gen 5 PRBS Gen 6 PRBS Rec 1 R/W R/W R/W R/W R/W R/W 0x3A 0x3B 0x3C 0x3D 0x3E 0x3F 0x00 0x00 0x00 0x00 0x00 0x00 0 0 0 PRBS Rec 2 PRBS Rec 3 R R 0x40 0x41 0x00 0x00 Reg Name R/W Readback/Status FREQMEAS0 R FREQMEAS1 R FREQMEAS2 R FREQ_RB1 R FREQ_RB2 R STATUSA R D7 D6 D5 FULLRATE LOS status D4 D3 D2 D1 FREQ0[7:0] (RATE_FREQ[7:0]) FREQ1[7:0] (RATE_FREQ[15:8]) FREQ2[7:0] (RATE_FREQ[23:16]) VCOSEL[7:0] DIVRATE[3:0] LOL status LOS done Static LOL CDR_MODE[2:0] 0 Reset static LOL VCOSEL[9:8] RATE_ MEAS_ COMP 0 LOL_ CONFIG LOS PDN LOS polarity 0 0 0 REFCLK_ PDN FREF_RANGE[1:0] D/PLL Control 0 EDGE_SEL[1:0] 0 0 0 0 D0 0 RATE_ MEAS_ EN 0 RATE_MEAS_ RESET 0 1 0 DATA_TO_REF_RATIO[3:0] TRANBW[2:0] ADAPTIVE_ DLL_SLEW[1:0] SLICE_EN SAMPLE_PHASE[3:0] Slice[6:0] INPUT_SEL[1:0] ADAPTIVE_ EQ_EN EQ_BOOST[3:0] SLICE_RB[7:0] 0 Data DATOUT_ CLKOUT_ 0 squelch DISABLE DISABLE DATA_SWING[3:0] LOS Control LOS_DATA[7:0] LOS_THRESHOLD[7:0] LOS_ LOS_ LOS_ 0 WRITE ENABLE RESET 0 DATA_ POLARITY CLOCK_SWING[3:0] CLOCK_ POLARITY LOS_ADDRESS[2:0] DATA_ DATA_GEN_ 0 CID_ EN EN DATA_CID_LENGTH[7:0] PROG_DATA[7:0] PROG_DATA[15:8] PROG_DATA[23:16] PROG_DATA[31:24] DATA_ DATA_ 0 RECEIVER_ RECEIVER_ ENABLE CLEAR PRBS_ERROR_COUNT[7:0] DATA_GEN_MODE[1:0] DATA_RECEIVER_ MODE[1:0] PRBS_ERROR Rev. 0 | Page 13 of 32 ADN2917 Reg Name PRBS Rec 4 PRBS Rec 5 PRBS Rec 6 PRBS Rec 7 ID/Revision REV ID HI_CODE LO_CODE Data Sheet R/W R R R R Addr (Hex) 0x42 0x43 0x44 0x45 Default (Hex) N/A N/A N/A N/A R R R R 0x48 0x49 0x20 0x21 0x54 0x15 0xFF 0xA6 D7 D6 D5 D4 D3 D2 DATA_LOADED[7:0] DATA_LOADED[15:8] DATA_LOADED[23:16] DATA_LOADED[31:24] D1 D0 REV[7:0] ID[7:0] Reserved Reserved Table 8. Status Register, STATUSA (Address 0x6) Bits D5 Bit Name LOS status D4 LOL status D3 LOS done D2 Static LOL D0 RATE_MEAS_COMP Bit Description 0 = no loss of signal 1 = loss of signal 0 = locked 1 = frequency acquisition mode 0 = LOS action not completed 1 = LOS action completed 0 = no LOL event since last reset 1 = LOL event since last reset; clear by using the static LOL bit, CTRLA[D2] 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 D7 D6:D4 Bit Name 0 CDR_MODE[2:0] D3 D2 D1 D0 0 Reset static LOL RATE_MEAS_EN RATE_MEAS_RESET Bit Description Reserved to 0. CDR modes. 001 = lock to data (LTD). 011 = lock to reference (LTR). 000, 010 = reserved. Reserved to 0. Set to 1 to clear static LOL. Fine data rate measurement enable. Set to 1 to initiate a rate measurement. Rate measurement reset. Set to 1 to clear a rate measurement. Table 10. Control Register, CTRLB (Address 0x9) Bits D7 D6 Bit Name SOFTWARE_RESET INIT_FREQ_ACQ D5 D4 0 LOL_CONFIG D3 LOS PDN D2 LOS polarity D1:D0 0 Bit Description Software reset. Write a 1 followed by a 0 to reset the device. Initiate frequency acquisition. Write a 1 followed by a 0 to initiate a frequency acquisition (optional). Reserved; CDR is always enabled. LOL configuration. 0 = normal LOL. 1 = static LOL. LOS power-down. 0 = normal LOS. 1 = LOS powered down. LOS polarity. 0 = active high LOS pin. 1 = active low LOS pin. Reserved to 0. Rev. 0 | Page 14 of 32 Data Sheet ADN2917 Table 11. Control Register, CTRLC (Address 0xA) Bits D7:D3 D2 D1 D0 Bit Name 0 REFCLK_PDN 0 1 Bit Description Reserved to 0. Reference clock power-down. Write a 0 to enable the reference clock. Reserved to 0. Reserved to 1. Table 12. Lock to Reference Clock Mode Programming Register, LTR_MODE 1 (Address 0xF) Bits D7 D6 Bit Name 0 LOL data D5:D4 FREF_RANGE[1:0] D3:D0 DATA_TO_REF_RATIO[3:0] Bit Description Reserved to 0 LOL data 0 = valid recovered clock vs. reference clock during tracking 1 = valid recovered clock vs. data during tracking fREF 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 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 D7:D5 D4:D3 Bit Name 0 EDGE_SEL[1:0] D2:D0 TRANBW[2:0] Bit Description Reserved to 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. Transfer bandwidth. Scales transfer bandwidth. Default value is 4. See the Transfer Bandwidth section for further details. Transfer BW = Default BW x (TRANBW[2:0]/4) Table 14. D/PLL Control Register, DPLLD (Address 0x13) Bits D7:D3 D2 D1:D0 Bit Name 0 ADAPTIVE_SLICE_EN DLL_SLEW[1:0] Bit Description Reserved to 0. Adaptive slice enable. 1 = enables automatic slice adjust. 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 D7:D4 D3:D0 Bit Name 0 SAMPLE_PHASE[3:0] Bit Description Reserved to 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 D7 Bit Name Extended slice D6:D0 Slice[6:0] Bit Description Extended slice enable. 0 = normal slice mode. 1 = extended slice mode. 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 D7 Bit Name RX_TERM_FLOAT D6:D5 INPUT_SEL[1:0] D4 ADAPTIVE_EQ_EN D3:D0 EQ_BOOST[3:0] Bit Description Rx termination float. 0 = termination common-mode driven. 1 = termination common-mode floated. Input stage select. 00: limiting amplifier. 01: equalizer. 10: 0 dB buffer. 11: undefined. Enable adaptive EQ. 0 = manual EQ control. 1 = adaptive EQ enabled. 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 D7:D6 D5 Bit Name 0 Data squelch D4 DATOUT_DISABLE D3 CLKOUT_DISABLE D2 D1 0 DATA_POLARITY D0 CLOCK_POLARITY Bit Description Reserved to 0 Squelch 0 = normal data 1 = squelch data Data output disable 0 = data output enabled 1 = data output disabled Clock output disable 0 = clock output enabled 1 = clock output disabled Reserved; double data rate is always enabled Data polarity 0 = normal data polarity 1 = flip data 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 D7:D4 Bit Name DATA_SWING[3:0] D3:D0 CLOCK_SWING[3:0] Bit Description 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. 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 compares the DCO frequency with the input data frequency. This third loop also sets the decimation ratio of the digital downsampler. 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. 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. 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. 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. 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. The delay-locked and phase-locked loops, together, simultaneously 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. 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 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. The phase of the input data signal is tracked by two separate feedback 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 PHASE-LOCKED LOOP (PLL) e(s) N BINARY PHASE DETECTOR KPLL x TRANBW Z(s) KDCO s I - z-1 /N RECOVERED CLOCK DELAY-LOCKED LOOP (DLL) N I - z-N KDLL I - z-1 I - z-1 PSH ZERO-ORDER HOLD SAMPLE CLOCK Z(s) X(s) = KPLL x TRANBW - KDCO s x N x PSH x KDLL + KPLL x TRANBW x KDCO Figure 18. CDR Jitter Block Diagram Rev. 0 | Page 18 of 32 11778-018 INPUT DATA X(s) 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 frequency 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 acquisition 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 to an on-chip voltage reference (VCM = 0.95 V typically). The inputs must be accoupled. 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 termination resistors absorb the difference in commonmode 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 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 mV. Table 20. Program Slice Level, Normal Slice Mode (Extended Slice = 0) Slice[6:0] 0000000 0000001 ... 1000000 ... 1111111 Decimal Value 0 1 ... 64 ... 127 Offset Slice function disabled -15 mV ... 0 mV ... +14.75 mV The amount of offset required for manual slice adjust is determined 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] 0000000 0000001 ... 1000000 ... 1111111 Decimal Value 128 129 ... 192 ... 255 Offset Slice function disabled -100 mV ... 0 mV ... +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. 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 DATA EDGE_SEL[1:0] 11778-019 CLK1 EDGE_SEL[1:0] = 2 CLK2 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 implementing 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 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: TRANBW[2:0] Transfer BW = ( Default Transfer BW ) x 4 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 1 2 3 4 5 6 7 Transfer BW (kHz) 350 700 1050 1400 (default) 1750 2100 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. 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 11778-021 DATA ADN2917 Data Sheet LOSS OF SIGNAL DETECTOR Signal Strength Measurement The receiver front-end LOS detector circuit detects when the input signal level falls below a user adjustable threshold. 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-topeak 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. 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 (2x) above the LOS threshold (see Figure 21). LOS OUTPUT INPUT VOLTAGE (VDIFF) PASSIVE EQUALIZER INPUT LEVEL HYSTERESIS t 11778-020 LOS THRESHOLD Figure 21. LOS Detector Hysteresis The LOS detector and the slice level adjust can be used simultaneously 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. 2. 3. 4. Write 0x21 to LOS_CTRL (Register 0x74). Write the desired threshold in millivolts to LOS_DATA (Register 0x36). Write 0x31 to LOS_CTRL (Register 0x74). Write 0x21 to LOS_CTRL (Register 0x74). 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) 6 10 15 20 to 30 Typical EQ Setting 10 12 14 15 Table 24. Typical EQ Reach on FR-4 vs. Maximum Data Rates Supported 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. Maximum Data Rate (Gbps) 4 8 10 11 I2C Register LOS_CTRL (Register 0x74) contains the necessary address and write enable bits to program this LOS threshold. Rev. 0 | Page 22 of 32 Typical EQ Reach on FR-4 (Inches) 30 20 15 10 Data Sheet ADN2917 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 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 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 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 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 resistors on the ADN2917. For example, the driver can be an open-drain switched current (see Figure 30), and the 50 resistors return this current to VCC. In Figure 30, the commonmode 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). 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. LOL 1 -1000 -250 0 250 1000 fDCO ERROR (ppm) 11778-022 0 dB EQ 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. 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 resistors in parallel, or 25 , so that the minimum current is 12 mA and the maximum current is 20 mA. For more details, see the Reference Clock (Optional) section. In this mode, the lock detector monitors the difference in frequency 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. LOCK DETECTOR OPERATION Static LOL Mode The lock detector on the ADN2917 has three modes of operation: normal mode, LTR mode, and 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 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 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 connected 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. Autoincrement 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 immediate 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 subaddress 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 series resistors present a differential load between REFCLKP and REFCLKN. Common mode is internally set to 0.56 x 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 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). ADN2917 REFCLKP 24 BUFFER 23 50 Table 25. LTR_MODE (Register 0xF) Settings 50 VCC/2 LTR_MODE[5:4] 00 01 10 11 11778-023 REFCLKN Figure 23. DC-Coupled, Differential REFCLKx Configuration ADN2917 VCC CLK OSC OUT REFCLKP 24 23 50 VCC/2 where DIV_fREF represents the divided-down reference referred to the 11.05 MHz to 22.1 MHz band. Figure 24. AC-Coupled, Single-Ended REFCLKx Configuration 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 ADN2917 REFCLKP 24 REFCLK BUFFER 23 50 50 VCC/2 9953.28 Mbps/19.44 MHz = 2(10 - 1) 11778-025 REFCLKN Ratio 2-1 20 2n - 1 29 DATA_TO_REF_RATIO[3:0] = data rate / DIV_fREF 11778-024 50 LTR_MODE[3:0] 0000 0001 n 1010 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 BUFFER REFCLKN Range (MHz) 11.05 to 22.1 22.1 to 44.2 44.2 to 88.4 88.4 to 176.8 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 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. 2. 3. 4. 5. 6. Apply the reference clock. Write a 0 to REFCLK_PDN (Bit D2 in Register 0xA) to enable the reference clock circuit. Write to FREF_RANGE[1:0] (Bits[D5:D4] in Register 0xF) to select the appropriate reference clock frequency circuit. 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. Write a 0-1-0 to RATE_MEAS_RESET (Bit D0 in Register 0x8). This initiates a new data rate measurement. 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: f DATARATE = (RATE _ FREQ [23:0 ] x f 2 LTR[5:4 ] 7 x2 x2 FULLRATE ) REFCLK DIVRATE x2 D15 to D8 FREQ1[7:0] 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 MeasurementTime = LSB D7 to D0 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 211 x 2 LTR[5:4] f REFCLK (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 approximately 5% without needing an external reference clock according to the following formula: (1) where: fDATARATE 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 D23 to D16 FREQ2[7:0] 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. Data = f DCO 2 FULLRATE x 2 DIVRATE (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] 0 1 2 3 ((65533) x (19.44 x 106))/(20 x 27 x 20 x 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 Rev. 0 | Page 26 of 32 Min Frequency (MHz) = Min_f(core) 5570 7000 8610 10,265 Max Frequency (MHz) = Max_f(core) 7105 8685 10,330 11,625 Data Sheet ADN2917 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: The following steps configure the PRBS generator: 1. fDCO = Min _ f (core) Max _ f (core) Min _ f (core) VCOSEL[7:0 ] 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 Table 27. PRBS Settings then PRBS Patterns PRBS7 PRBS15 PRBS31 PROG_DATA [31:0]1 fDCO = 10265 Mbps (11625 10265 ) Mbps 256 17 10355 .31 Mbps and f data 10355.31 Mbps 20 20 2. 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. 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 x DATA_CID_LENGTH[7:0] (Bits[D7:D0] in Register 0x3A), which is set via PRBS Gen 2[7:0] register (Register 0x3A). 10355.31 Mbps 1 DATA_GEN_MODE[1:0] (Bits[D1:D0] in Register 0x39) 0x00 0x01 0x10 0x11 PRBS Polynomial 1 + X6 + X7 1 + X14 + X15 1 + X28 + X31 Not applicable Bits[D7:D0] in Register 0x3B, Register 0x3C, Register 0x3D, and Register 0x3E. Initiate Frequency Acquisition Double Data Rate Mode 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). The recovered output clock is a double data rate (DDR) clock, where the output clock frequency is 1/2 the data rate. This allows direct interfacing to FPGAs that support clocking on both rising and falling edges. Disable Output Buffers The following steps configure the PRBS detector: 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. 1. Transmission Lines 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. 2. 3. 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). 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). 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. Use of 50 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 terminated with 50 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. Rev. 0 | Page 27 of 32 ADN2917 Data Sheet 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. The high speed outputs (DATOUTP, DATOUTN, CLKOUTP, and CLKOUTN) are internally terminated with 50 to VCC. Soldering Guidelines for Lead Frame Chip Scale Package INPUT CONFIGURATIONS 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. 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. SLICE ADJUST 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. LOS DETECT LOS LA PIN 2 BYPASS NIN 2.9k 2.9k 50 50 EQ INPUT_SEL[1:0] 11778-026 RX_TERM_FLOAT VCC VREF FLOAT Figure 26. Input Stage Circuit Block Diagram Table 28. Input Signal Configuration Selected Input Limiting Amplifier Equalizer Bypass (0 dB Buffer) Not Defined INPUT_SEL[1:0] (Bits[D6:D5] in Register 0x16) 00 01 10 11 Rev. 0 | Page 28 of 32 RX_TERM_FLOAT (Bit D7 in Register 0x16) = 0 VREF VREF VCC Not defined RX_TERM_FLOAT = 1 Not defined Not defined Float Not defined Data Sheet ADN2917 PACKAGE EXPOSED PAD PCB PAD WITH 25 VIAS COPPER PLANE--V EE HEAT DISSIPATION HEAT DISSIPATION 11778-027 COPPER PLANE--V EE 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 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 capacitor 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%. where: n is the number of CIDs. T is the bit period. Calculate the capacitor value by combining the equations for 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. Normalizing to V p-p, Droop = V = 0.04 V = 0.5 V p-p (1 - e-t/) = 12t where: is the RC time constant (C is the ac coupling capacitor, R = 100 seen by C). t is the total discharge time t = n Rev. 0 | Page 29 of 32 ADN2917 Data Sheet VCC V1 V2 ADN2917 PIN DATOUTP 50 TIA VREF CIN 2 CDR COUT 50 V1b 1 V2b 2 DATOUTN NIN 3 4 V1 V1b V2 VREF V2b VDIFF VTH VDIFF = V2 - V2b VTH = ADN2917 QUANTIZER THRESHOLD 2. WHEN THE TIA OUTPUTS CONSECUTIVE IDENTICAL DIGITS, V1 AND V1b ARE DRIVEN TO DIFFERENT DC LEVELS. V2 AND V2b DISCHARGE TO THE VREF LEVEL, WHICH EFFECTIVELY INTRODUCES A DIFFERENTIAL DC OFFSET ACROSS THE AC COUPLING CAPACITORS. 3. WHEN THE BURST OF DATA STARTS AGAIN, THE DIFFERENTIAL DC OFFSET ACROSS THE AC COUPLING CAPACITORS IS APPLIED TO THE INPUT LEVELS, CAUSING A DC SHIFT IN THE DIFFERENTIAL INPUT. THIS SHIFT IS LARGE ENOUGH SUCH THAT ONE OF THE STATES, EITHER HIGH OR LOW, DEPENDING ON THE LEVELS OF V1 AND V1b WHEN THE TIA BEGAN DETECTING AND OUTPUTTING A CID DATA SYSTEM, 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. Figure 28. Example of Baseline Wander Rev. 0 | Page 30 of 32 11778-028 NOTES 1. DURING THE DATA PATTERNS WITH HIGH TRANSITION DENSITY, DIFFERENTIAL DC VOLTAGE AT V1 AND V2 IS ZERO. Data Sheet ADN2917 VCC DC-COUPLED APPLICATION ADN2917 50 50 PIN 50 NIN 50 50 VCC 11778-031 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. I Figure 31. DC-Coupled Application, 0 dB EQ Input (Back Terminated Mode) ADN2917 VDD PIN TIA TIA 1.2V 50 0.8V NIN 50 VCM = 1.05V INPUT (V) Figure 29. DC-Coupled Application, 0 dB EQ Input (Rx Term Float Mode) VCM = 0.65V 600mV p-p, DIFF 600mV p-p, DIFF 0.9V 0.5V 11778-032 11778-029 50 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. Figure 32. Minimum Allowed DC-Coupled Input Levels 1.2V ADN2917 1.0V PIN VCM = 0.95V INPUT (V) VCM = 0.75V 50 NIN 50 0.5V 11778-030 VCC I 0.7V 11778-033 50 1.0V p-p, DIFF 1.0V p-p, DIFF Figure 33. Maximum Allowed DC-Coupled Input Levels Figure 30. DC-Coupled Application, 0 dB EQ Input (Normal Mode) Rev. 0 | Page 31 of 32 ADN2917 Data Sheet OUTLINE DIMENSIONS 4.10 4.00 SQ 3.90 0.60 MAX 2.50 REF 0.60 MAX 3.75 BSC SQ 1 0.50 BSC 2.65 2.50 SQ 2.35 EXPOSED PAD 6 13 0.50 0.40 0.30 TOP VIEW 1.00 0.85 0.80 12 MAX 0.80 MAX 0.65 TYP 0.05 MAX 0.02 NOM COPLANARITY 0.08 0.20 REF 0.30 0.23 0.18 SEATING PLANE 7 12 BOTTOM VIEW 0.25 MIN FOR PROPER CONNECTION OF THE EXPOSED PAD, REFER TO THE PIN CONFIGURATION AND FUNCTION DESCRIPTIONS SECTION OF THIS DATA SHEET. COMPLIANT TO JEDEC STANDARDS MO-220-VGGD-8 04-11-2012-A PIN 1 INDICATOR PIN 1 INDICATOR 24 19 18 Figure 34. 24-Lead Lead Frame Chip Scale Package [LFCSP_VQ] 4 mm x 4 mm Body, Very Thin Quad (CP-24-3) Dimensions shown in millimeters ORDERING GUIDE Model1 ADN2917ACPZ ADN2917ACPZ-RL7 EVALZ-ADN2917 1 Temperature Range -40C to +85C -40C to +85C Package Description 24-Lead LFCSP_VQ 24-Lead LFCSP_VQ Package Option CP-24-3 CP-24-3 Z = RoHS Compliant Part. I2C refers to a communications protocol originally developed by Philips Semiconductors (now NXP Semiconductors). (c)2014 Analog Devices, Inc. All rights reserved. Trademarks and registered trademarks are the property of their respective owners. D11778-0-5/14(0) Rev. 0 | Page 32 of 32 Ordering Quantity 490 1500 Mouser Electronics Authorized Distributor Click to View Pricing, Inventory, Delivery & Lifecycle Information: Analog Devices Inc.: ADN2917ACPZ EVALZ-ADN2917 ADN2917ACPZ-RL7