PCIe 4.0 Physical Layer Transmitter and Receiver Testing April 2017 Rick Eads PCI Express Solutions Planner Page
Agenda PCIe 4.0 Ecosystem and Timeline PCIe 4.0 TX Testing and Tools PCIe U.2 Testing RX Testing and Link/EQ PCIe 4.0 16GT/s RX Testing Page 2
PCIe Ecosystem PCISIG Board of Directors Intel, AMD, IBM, Synopsys, Qualcomm, Dell, HP, NVIDIA, Lenovo Electrical Work Group Protocol Work Group Card Electromechanic al Work Group Serial Enabling Work Group Deliverables: Group Chairs: Electrical Spec AMD, Intel Protocol Spec AMD, Intel CEM Spec Intel Test Specification & Plugfests Intel, Synopsys PCI Express 4.0 PCIe 4.0 16GT/s TX/RX Testing Page 3
PCIe Ecosystem PCISIG Owned FFs U.2 (SFF-8639) M.2 (SATA, USB, PCI-E) Intel, AMD, IBM, Synopsys, Qualcomm, Dell, HP, NVIDIA, Lenovo PCISIG Owned Oculink Cabled PCIe mpci (MIPI.org) Non-PCISIG Owned CCIX SATA Express NVMe Electrical Work Group Protocol Work Group Card Electromechanic al Work Group Serial Enabling Work Group Deliverables: Group Chairs: Electrical Spec AMD, Intel Protocol Spec AMD, Intel CEM Spec Intel Test Specification & Plugfests Intel, Synopsys PCI Express 4.0 PCIe 4.0 16GT/s TX/RX Testing Page 4
PCI Express 4.0 Timeline (estimated) Rev 0.3 Rev 0.5 Rev 0.7 Rev 0.9 Q1 Q2 Q3 Q4 Q1 Q2 Q3 Q4 Q1 Q2 Q3 Q4 Q1 Q2 Q3 Q4 Q1 Q2 Q3 Q4 2013 2014 2015 2016 2017 PCIe 4.0 16GT/s TX/RX Testing Page 5
PCI Express Specifications and Scope Select the specifications that relate to your need Base Specification Contains all the system knowledge Can directly be applied to Chip Test Card Electromechanical (CEM) Spec Applies to Add-In Cards and Mother Boards Mitigates card manufacturer s need to study the base specification Increases reproducibility through PCI-SIG supplied test tools CBB and CLB (compliance base and load board) Phy Test Specification Defines compliance tests of CEM spec in detail PCIe 4.0 16GT/s TX/RX Testing Page 6
PCI Express Specifications and Scope Select the specifications that relate to your need Base Specification Contains all the system knowledge Can directly be applied to Chip Test Soon to Be Released at v0.7 Card Electromechanical (CEM) Spec Applies to Add-In Cards and Mother Boards Mitigates card manufacturer s need to study the base specification Increases reproducibility through PCI-SIG supplied test tools CBB and CLB (compliance base and load board) Phy Test Specification Defines compliance tests of CEM spec in detail PCIe 4.0 16GT/s TX/RX Testing Page 7
PCIe 4.0 New Features Based on PCIe v0.7 BASE specification New data rate:16gt/s Requires an output stages capable of providing pre-shoot and de-emphasis with fast enough risetimes. Link Equalization protocol similar to PCIe 3.0 TxEQ P0-P10 RxEQ CTLE (2 pole 1 zero) + 2tap DFE Max Channel Length -28dB @ 8GHz & 1 connector Re-timers used for longer channels or for channels with >1 connector RX clocking architectures: CC and IR CC -> Common RefClock -> synchronous RX and TX w/ or w/o SSC IR -> Independent RefClock -> asynchronous RX and TX w/ or w/o SSC Initial LinkEQ speed selection: 2.5GT/s -> 8GT/s with link equalization if successful -> Then transitions to 16GT/s with another round of link equalization TX Jitter Analysis: Similar to PCIe 3.0 Lane Margining added. PCIe 4.0 16GT/s RX Testing Page 8
PCIe PHY-layer Specification Differences PCIe 1.1, 2.0, 3.0, and 4.0 PCIe 1 PCIe2 PCIe3 PCIe4 transfer rate 2.5GT/s 5GT/s 8GT/s 16GT/s coding 8B/10B 8B/10B 128B/130B 128B/130B overhead 20% 20% 1.5625% 1.5625% symbol / block alignment scrambling K28.5 for symbol alignment optional with PRBS 2 16-1; scrambler reset through K28.5 K28.5 for symbol alignment optional with PRBS 2 16-1; scrambler reset through K28.5 EIEOS for block alignment control: no (partially), data: always PRBS 2 23-1; scrambler reset through EIEOS EIEOS for block alignment data rate / lane 2 GB/s 4Gb/s 7.875Gb/s 15.75Gb/s control: no (partially), data: always PRBS 2 23-1; scrambler reset through EIEOS Equalization TX: -3.5dB RX: None TX: -3.5dB, -6dB RX: None TX: 2 Tap FIR RX: CTLE + 1 Tap DFE TX: 2 Tap FIR RX: CTLE + 2 Tap DFE TX-test normative (required) normative (required) normative (required) normative (required) PCIe 4.0 16GT/s RX Testing Page 9
CEM 4.0 and Compliance Testing CEM 4.0 currently at v0.5. V0.7 in CEM Review (Jan 2017) PCIe 4.0 Compliance Requirements CEM Spec completion at v0.7 (v0.9 optimal) Completion of Test Specifications Config Test Spec Link Transaction Test Spec System Firmware (BIOS) Test Spec Electrical Test Spec Retimer Test Spec Availability of Gen4 Compliance Test Fixtures for Purchase Preliminary PCIe 4.0 Test Fixtures to debut at April 2017 Workshop Estimated Schedule First Gen4 FYI testing commences April 2017 Official FYI Testing to begin late 2017 Official Integrators list test to start mid 2018 PCIe 4.0 16GT/s TX/RX Testing Page 10
PCI Express 4.0 Keysight Total Solution Physical layer interconnect design Physical layertransmitter test Physical layerreceiver test ADS design software 90000 X, Z-Series oscilloscope M8020A J-BERT High Perfformance BERT 86100D DCA-J/TDR N5393F PCI Express electrical compliance software N5990A automated compliance and device characterization test software E5071C ENA option TDR 86100CU-400 PLL and Jitter Spectrum Measurement SW Industry s lowest scope noise floor/sensitivity and trigger jitter DSA-X Series & Q Series Real-Time Oscilloscopes Automated compliance software accurate, efficient and consistent PCIe 4.0 16GT/s TX/RX Testing Page 11
PCIe 4.0 TX Testing PCIe 4.0 16GT/s TX/RX Testing Page 12
PCIe 4.0 v0.9 Finalizes TX Scope Bandwidth Requirement Gen4 De-embed limit Gen4 Scope BW limit PCIe 4.0 16GT/s RX Testing Page 13
PCI Express 4.0 TX Measurement Basic Test Setup BASE Spec (v0.7) Keysight Z-Series Real Time Oscilloscope PCIe 4.0 ASIC/IC Custom Breakout Board S-Parameters of Replica Ch. Used to de-embed to pin. PCIe 4.0 16GT/s RX Testing Page 14
N5393F/G New Features Supports PCIe 4.0 BASE TX Testing at 2.5G, 5G, 8G and 16GT/s (v0.7 BASE) Supports PCIe 4.0 Reference Clock tests (2.5G, 5G, 8G, 16G) Supports U.2 (SFF-8639) CEM tests for endpoints and root complexes (2.5G, 5G, 8G). Automated DUT control using an 81150/60A Pulse Generator ARB. Enhanced Switch Matrix supporting arbitrary lane mapping New Workshop Compliance Mode for rapid PCISIG official compliance testing. PCIe 4.0 16GT/s RX Testing PCIe 4.0 16GT/s TX Testing Page 15
Keysight PCIe 4.0 (Gen4) TX N5393F Test Application New Test Plan Setup Select Speeds of Gen4 Device to Test Select Standard Version to Test Automatic DUT control for toggle signal PCIe 4.0 16GT/s TX Testing Page 16
Keysight PCIe 4.0 (Gen4) TX N5393F Test Application Select a complete Gen4 TX test plan. PCIe 4.0 16GT/s TX Testing Page 17
Keysight PCIe 4.0 (Gen4) TX N5393F Test Application Use InfiniiSim for de-embedding with optional N5465A Select InfiniiSim under Tools Choose your de-embed transfer function Fine-tune your de-embed filter (bandwidth, etc) PCIe 4.0 16GT/s TX Testing Page 18
Consolidated Jitter Parameters for ALL data rates 19 Note Gen1, Gen2 Jitter now has same measurement parameter as Gen3/4 but different values. Symbol V TX-BOOST-RS EQ TX-COEFF-RES V TX-DE-RATIO- 3.5dB V TX-DE-RATIO-6dB T TX-UTJ T TX-UTJ-SRIS T TX-UDJDD T TX-UPW-TJ T TX-UPWDJDD Maximum nominal Tx boost ratio for reduced swing Tx coefficient resolution Tx de-emphasis ratio for 2.5 and 5 G Tx de-emphasis ratio for 5 G Tx uncorrelated total jitter Tx uncorrelated total jitter when testing for the IR clock mode with SSC Tx uncorrelated Dj for non-embedded Refclk Total uncorrelated pulse width jitter Deterministic DjDD uncorrelated pulse width jitter T TX-RJ Tx Random jitter N/A L TX-SKEW RL TX-DIFF RL TX-CM Parameter description Lane-to-Lane Output Skew Tx package plus die differential return loss Tx package plus die common mode return loss 2.5 GT/s 5.0 GT/s 8.0 GT/s 16.0 GT/s Units Notes N/A N/A 2.5 ~2.5 (min) db Assumes ±1.0 db tolerance from diagonal elements in Table 9-3. N/A N/A 1/24 (max) 1/63 (min) 3.0 (min) 4.0 (max) 3.0 (min) 4.0 (max) 1/24 (max) 1/63 (min) N/A N/A N/A db N/A 5.5 (min) 6.5 (max) N/A N/A db 100 (max) 50 (max) 31.25 (max) 12.5 (max) ps PP at 10-12 100 (max) 66.51 (max) 33.83 (max) 15.85 (max) ps PP at 10-12 See Sec 9.3.5.8 for details. See Sec 9.3.5.8 for details. 100 (max) 30 (max) 12 (max) 6.25 (max) ps PP See Sec 9.3.5.8 for details. N/A 40 (max) 24 (max) 12.5 (max) ps PP at 10-12 See Sec 9.3.5.9 for details N/A 40 (max) 10 (max) 5 (max) ps PP See Sec 9.3.5.9 for details 1.4 2.2.45 -.89 ps RMS Informative 1.4-3.6 parameter only. Range of Rj possible with zero to maximum CDR SRIS 8GG, allowed 16G T TX- UDJDD. 2.5 (max) 2.0 (max) 1.5 (max) 1.25 (max) ns Between any two Lanes within a single Transmitter. See Figure 9-19 See Figure 9-20 See Figure 9-19 See Figure 9-20 See Figure 9-19 See Figure 9-20 See Figure 9-19 See Figure 9-20 db Note 6 db Note 6 Page
PCIe 4.0 Reference Clock Testing PCIe 4.0 16GT/s TX Testing Page 20
PCIe 4.0 Reference Clock AC Parameters Symbol Table 2-1: REFCLK DC Specifications and AC Timing Requirements Parameter 100 MHz Input Rising Edge Rate Rising Edge Rate 0.6 4.0 V/ns 2, 3 Falling Edge Rate Falling Edge Rate 0.6 4.0 V/ns 2, 3 VIH Differential Input High Voltage +150 mv 2 VIL Differential Input Low Voltage -150 mv 2 VCROSS Absolute crossing point voltage +250 +550 mv 1, 4, 5 VCROSS DELTA Variation of VCROSS over all rising clock edges +140 mv 1, 4, 9 VRB Ring-back Voltage Margin -100 +100 mv 2, 12 TSTABLE Time before VRB is allowed 500 ps 2, 12 TPERIOD AVG Average Clock Period Accuracy -300 +2800 ppm 2, 10, 13 TPERIOD ABS Absolute Period (including Jitter and Spread Spectrum modulation) 9.847 10.203 ns 2, 6 TCCJITTER Cycle to Cycle jitter 150 ps 2 VMAX Absolute Max input voltage +1.15 V 1, 7 VMIN Absolute Min input voltage - 0.3 V 1, 8 Duty Cycle Duty Cycle 40 60 % 2 Rising edge rate (REFCLK+) to Rise-Fall falling edge rate (REFCLK-) 20 % 1,14 Matching matching ZC-DC Clock source DC impedance 40 60 1, 11 Min Max Unit Note 21 Standard Reference Clock Test Fixture for ALL data rate independent and data rate dependent parameters. PCIe 4.0 16GT/s TX Testing Page
PCIe 4.0 Phase Jitter Parameters 9.6.7 Jitter Limits for Refclk Architectures Table 9-18 lists the jitter limits for the CC Refclk architecture at each of the four data rates Jitter at 2.5 GT/s is measured as a peak to peak jitter value, because a substantial proportion of the jitter is SSC harmonics which appears at the receiver as Dj. The combination of the 2.5 GT/s PLL and CDR bandwidths passes a significant amount of SSC residual, where it appears Dj. The 108 ps number is the same as that specified in the 3.0 CEM spec. For 5.0, 8.0, and 16.0 GT/s jitter is specified as an RMS (Rj) value. These signaling speeds utilize a lower PLL BW and a higher CDR BW, and the effect is to suppress SSC harmonics such that almost all the jitter appears as Rj. Data Rate CC jitter Limit Notes 2.5G 108 ps pp 1, 2 5.0G 3.1 ps RMS 1, 2 8.0G 1.0 ps RMS 1, 2 16G Table 9-18: Jitter Limits for CC Architecture 0.5 ps RMS Note that.7 ps RMS is to be used in channel simulations to account for additional noise in a real system. Note: 1. The Refclk jitter is measured after applying the filter function in Figure 9-45 2. Jitter measurements shall be made with a capture of at least 100,000 clock cycles captured by a real time oscilloscope with a sample rate of 20 GS/s or great. Broadband oscilloscope noise must be minimized in the measurement. 1,2 Standard Reference Clock Test Fixture for Phase Jitter Parameters 22 PCIe 4.0 16GT/s TX Testing Page
Reference Clock for Gen4 PLL 1 PLL 2 Table 9-17: Common Refclk PLL and CDR Characteristics for 8.0 and 16.0 GT/s Show Spec in Report PLL #1 PLL #2 0.01 db peaking 1.0 db peaking 0.01 db peaking 3.0 db peaking BW PLL (min) = 5.0 MHz BW PLL (max) = 16 MHz w n1 = 1.12 Mrad/s z 1 = 14 w n1 = 3.58 Mrad/s z 1 = 14 w n1 = 11.01 Mrad/s z 1 = 1.16 w n1 = 35.26 Mrad/s z 1 = 1.16 BW PLL (min) = 8.0 MHz BW PLL (max) = 16 MHz w n2 = 1.79 Mrad/s z 2 = 14 w n2 = 3.58 Mrad/s z 2 = 14 w n2 = 26.86 Mrad/s z 2 = 0.54 w n2 = 53.73 Mrad/s z 2 = 0.54 BW CDR (min) = 5 MHz, 1 st order 64 combinations 5 GT/s REF Clock 8G TX Phase Jitter PLL1 PLL 2 ATX BTX CTX DTX ATX BTX CTX DTX ARX BRX CRX 0.3 ps 7.9ps 0.99 DRX ARX BRX CRX DRX If user right clicks on a curve, pop up menu shows curves related to the calculation of that value. Color code fields for PASS/Fail/Margin PCIe 4.0 16GT/s TX/RX Testing Report Results in a Matrix Page 23
Keysight PCIe 4.0 (Gen4) TX N5393F Test Application Gen4 Reference Clock Tests Common 100MHz AC Parameters Phase Jitter tests for 2.5G, 5G, 8G, and 16G SSC Clock Tests PCIe 4.0 16GT/s TX Testing Page 24
PCIe 3.0 U.2 Testing (SFF-8639) PCIe 4.0 16GT/s TX Testing Page 25
The U.2 Connector Pinout PCIe 4.0 16GT/s TX Testing Page 26
New U.2 CBB Fixtures to be made available from PCI-SIG PCIe 4.0 16GT/s TX Testing Page 27
Data and Clock Connections to U.2 connector on Backplane Refclk Connection Data Connection 28 PCIe 4.0 16GT/s TX Testing Page
Keysight PCIe 4.0 (Gen4) TX N5393F Test Application U.2 Test Setup U.2 Test Selection for Endpoint Card PCIe 4.0 16GT/s RX Testing Page 29
Speeding up Testing Automated DUT Control Enhanced Switch Matrix Lane Mapping Workshop Compliance Mode PCIe 4.0 16GT/s RX Testing PCIe 4.0 16GT/s TX Testing Page 30
Keysight PCIe 4.0 (Gen4) TX N5393F*Test Application For use with PCIe 3.0 and below. Use and 81150A or 81160A to Generate the CBB Compliance Toggle signal to toggle your DUT between the different compliance states. Control DUT automatically to switch compliance toggle modes Workshop compliance mode is used for PCISIG compliance testing and uses Sigtest to test your DUT as well as to create PCISIG Compliance test reports. You specify what directory to use for your Workshop Compliance Mode (Sigtest generated) HTML reports along with data files PCIe 4.0 16GT/s RX Testing Page 31
Keysight PCIe 4.0 (Gen4) TX N5393F*Test Application Connection Options Choose from available switch matrix options for multi-lane testing If you don t have a switch, you can test using all four scope channels to test two lanes with one setup. Select Lanes to map to your switch network setup Lanes to test can be chosen arbitrarily. PCIe 4.0 16GT/s RX Testing Page 32
Keysight PCIe 4.0 (Gen4) TX N5393F*Test Application Test Automation and Connection Example 81150A used for DUT control (toggle pulse to switch compliance states) Scope configured for Root Complex Testing Root Complex DUT being tested for PCISIG Compliance Keysight U3020A Switch Matrix PCIe 4.0 16GT/s RX Testing Page 33
PCIe 4.0 Receiver Testing at 16Gbps >16Gbps BERT Link Equalization Jitter and De-emphasis PCIe 4.0 16GT/s RX Testing Page 34
Differences between PCIe 3 and PCIe 4 PCIe 3.0/3.1 PCIe 4.0 rev 0.5 PCIe 4.0 rev 0.7 added transfer rate 8 GT/s 16 GT/s coding block alignment & scrambler reset 128B/130B EIEOS for block alignment EIEOS 10 00FF 00FF 00FF 00FF 00FF 00FF 00FF 00FF 500 MHz scrambling 10 00FF 00FF 00FF 00FF 00FF 00FF 00FF 00FF 1 GHz control: no (partially), data: always PRBS 2 23-1; scrambler reset through EIEOS 10 0000 FFFF 0000 FFFF 0000 FFFF 0000 FFFF 500 MHz Adaptable TX link equalization yes yes, two step process: first 8G link eq followed by 16G link eq if 8G link eq is successful RX tests stressed jitter test and stressed voltage test one RX stress test rssc for common reference clock no no yes eye opening after reference RX for stress signal cal 0.3 UI, 25 mv, BER of 10-12 0.3 UI, 15 mv (RX eye spec. is actually 14 mv), BER of 10-12 stress signal adjustment using RJ, DM-SI and V diff coarse: ISI fine: DM-SI + SJ or DM-SI + V diff CTLE pole 1 frequency 2 GHz 4 GHz 2 GHz Channel for RX test No connector required PCIe 4.0 CEM connector required as part of RX test channel Page
Differences between PCIe 3 and PCIe 4 Relevant changes with PCIe 4.0 rev 0.5 and 0.7 PCIe 3.0/3.1 PCIe 4.0 rev 0.5 PCIe 4.0 rev 0.7 added transfer rate 8 GT/s 16 GT/s coding block alignment & scrambler reset 128B/130B EIEOS for block alignment EIEOS 10 00FF 00FF 00FF 00FF 00FF 00FF 00FF 00FF 500 MHz scrambling Link EQ gets more important 10 00FF 00FF 00FF 00FF 00FF 00FF 00FF 00FF 1 GHz control: no (partially), data: always PRBS 2 23-1; scrambler reset through EIEOS 10 0000 FFFF 0000 FFFF 0000 FFFF 0000 FFFF 500 MHz Adaptable TX link equalization yes yes, two step process: first 8G link eq followed by 16G link eq if 8G link eq is successful RX tests stressed jitter test and stressed voltage test one RX stress test rssc for common reference clock no no yes eye opening after reference RX for stress signal cal Different cal procedure 0.3 UI, 25 mv, BER of 10-12 0.3 UI, 15 mv (RX eye spec. is actually 14 mv), BER of 10-12 stress signal adjustment using RJ, DM-SI and V diff coarse: ISI fine: DM-SI + SJ or DM-SI + V diff Reference CTLE changes: pole 1 frequency affects RX cal Special cal channel fixture required 2 GHz 4 GHz 2 GHz Channel for RX test No connector required PCIe 4.0 CEM connector required as part of RX test channel Page
Differences between PCIe 3 and PCIe 4 Relevant changes between PCIe 4.0 rev 0.5 to rev 0.7 PCIe 3.0/3.1 PCIe 4.0 rev 0.5 PCIe 4.0 rev 0.7 added transfer rate 8 GT/s 16 GT/s coding Affects block alignment and descrambler reset block alignment & scrambler reset 128B/130B EIEOS for block alignment EIEOS 10 00FF 00FF 00FF 00FF 00FF 00FF 00FF 00FF 500 MHz scrambling A new type of impairment. Amplitude is significantly higher compared to PCIe 2.0 10 00FF 00FF 00FF 00FF 00FF 00FF 00FF 00FF 1 GHz control: no (partially), data: always PRBS 2 23-1; scrambler reset through EIEOS 10 0000 FFFF 0000 FFFF 0000 FFFF 0000 FFFF 500 MHz Adaptable TX link equalization yes yes, two step process: first 8G link eq followed by 16G link eq if 8G link eq is successful RX tests stressed jitter test and stressed voltage test one RX stress test rssc for common reference clock no no yes Impacts RX stress test calibration eye opening after reference RX for stress signal cal 0.3 UI, 25 mv, BER of 10-12 0.3 UI, 15 mv (RX eye spec. is actually 14 mv), BER of 10-12 stress signal adjustment using RJ, DM-SI and V diff coarse: ISI fine: DM-SI + SJ or DM-SI + V diff Reference CTLE changes: pole 1 frequency 2 GHz 4 GHz 2 GHz Channel for RX test No connector required PCIe 4.0 CEM connector required as part of RX test channel Page
Blub AIC Test System Test EQ starts EQ complete Dynamic Link Equalization Handshake 8G The four phases of the Link Equalization Protocol RcvrLock Phase 0 Phase 1 Phase 2 UPSTREAM PORT Add-In Card J-BERT M8020A TS1, [P2] EC = 00b, PV = P2 TS1, [P2] EC = 01b, PV = P2, Use_Preset = 0 TS1, [P2] EC = 10b, PV = P3, Use_Preset = 1 TS1, [P2] EC = 10b, PV = P4, Use_Preset = 1 BER < 10-12 Phase 3 TS1, [P2] EC = 11b, PV = P2, Use_Preset = 0 TS1, [P6] EC = 11b, PV = P6, Use_Preset = 0 TS1, [P7] EC = 11b, PV = P7, Use_Preset = 0 RcvrLock TS1, [P7] EC = 00b, PV = P7 EQTS2 PV = P1 DOWNSTREAM PORT J-BERT M8020A System TS1, [P1] EC = 01b, PV = P1, Use_Preset = 0 TS1, [P1] EC = 10b, PV = P1, Use_Preset = 0 TS1, [P3] EC = 10b, PV = P3, Use_Preset = 0 TS1, [P4] EC = 10b, PV = P4, Use_Preset = 0 TS1, [P4] EC = 11b, PV = P6, Use_Preset = 1 TS1, [P4] EC = 11b, PV = P7, Use_Preset = 1 TS1, [P4] EC = 00b, PV = P4, Use_Preset = 0 RcvrLock Phase 1 Phase 2 Phase 3 BER < 10-12 RcvrLock PV EC Phase 0: 2.5 Gb/s Downstream port tells upstream port which initial preset to use after the speed change will have been done. Phase 1: 8 Gb/s Link partners settle on 8 GT/s speed. Exchange FS/LF values. Phase 2: 8 Gb/s Add-in Card sets up the deemphasis of the System Board s transmitter. Phase 3: 8 Gb/s System Board sets up the deemphasis of the Add-in Card s transmitter. Preset Value Equalization Control PCIe 4.0 16GT/s RX Testing Page 38
Blub AIC Test System Test EQ starts EQ complete Dynamic Link Equalization Handshake 16G The four phases of the Link Equalization Protocol RcvrLock Phase 0 Phase 1 Phase 2 UPSTREAM PORT Add-In Card J-BERT M8020A TS1, [P2] EC = 00b, PV = P2 TS1, [P2] EC = 01b, PV = P2, Use_Preset = 0 TS1, [P2] EC = 10b, PV = P3, Use_Preset = 1 TS1, [P2] EC = 10b, PV = P4, Use_Preset = 1 BER < 10-12 Phase 3 TS1, [P2] EC = 11b, PV = P2, Use_Preset = 0 TS1, [P6] EC = 11b, PV = P6, Use_Preset = 0 TS1, [P7] EC = 11b, PV = P7, Use_Preset = 0 RcvrLock TS1, [P7] EC = 00b, PV = P7 EQTS2 PV = P1 DOWNSTREAM PORT J-BERT M8020A System TS1, [P1] EC = 01b, PV = P1, Use_Preset = 0 TS1, [P1] EC = 10b, PV = P1, Use_Preset = 0 TS1, [P3] EC = 10b, PV = P3, Use_Preset = 0 TS1, [P4] EC = 10b, PV = P4, Use_Preset = 0 TS1, [P4] EC = 11b, PV = P6, Use_Preset = 1 TS1, [P4] EC = 11b, PV = P7, Use_Preset = 1 TS1, [P4] EC = 00b, PV = P4, Use_Preset = 0 RcvrLock Phase 1 Phase 2 Phase 3 BER < 10-12 RcvrLock PV EC Phase 0: 2.5 Gb/s Downstream port tells upstream port which initial preset to use after the speed change will have been done. Phase 1: 8 Gb/s Link partners settle on 8 GT/s speed. Exchange FS/LF values. Phase 2: 8 Gb/s Add-in Card sets up the deemphasis of the System Board s transmitter. Phase 3: 8 Gb/s System Board sets up the deemphasis of the Add-in Card s transmitter. Preset Value Equalization Control IF SUCCESSFUL Phase 1: 16 Gb/s Link partners settle on 16 GT/s speed. Exchange FS/LF values. Phase 2: 16 Gb/s Add-in Card sets up the deemphasis of the System Board s transmitter. Phase 3: 16 Gb/s System Board sets up the deemphasis of the Add-in Card s transmitter. PCIe 4.0 16GT/s RX Testing Page 39
PCIe 4.0 RX Tests 16 GT/s Base Specification RX Test Setup PCIe 4.0 Base Spec requires a CEM connector to be part of the test channel! All other impairments are provided by J-BERT M8020A A built-in reference clock multiplier enables J-BERT M8020A to operate on a DUT s reference clock if required No ref clock connection in case of IR / SRIS PCIe 4.0 16GT/s RX Testing Page 40
PCIe 4.0 Calibration Channel Example (PCISIG) CEM Connector. 85ohm Differential Impedance PCIe 4.0 16GT/s TX/RX Testing Page 41
PCIe 4.0 RX Tests 16 GT/s Base Specification RX SJ Mask Different jitter tolerance curves for CC and IR rssc for CC only Constant 33 khz SJ tone in addition to jitter tolerance curve and 210 MHz SJ tone for IR only PCIe 4.0 16GT/s RX Testing Page 42
PCIe 4.0 RX Tests Test Automation Support for CC as well as IR End point as well as root complex 2.5 GT/s, 5 GT/s, 8 GT/s and 16 GT/s Contact Keysight for availability PCIe 4.0 16GT/s RX Testing Page 43
PCIe 4.0 RX Stress Signal Calibration 16 GT/s Receiver Stress Signal Calibration Setup 1 V diff, pre-shoot and de-emphasis calibration N5990A Test Automation SW for PCIe RJ calibration SJ calibration PCIe 4.0 16GT/s RX Testing Page 44
PCIe 4.0 RX Stress Signal Calibration 16 GT/s Receiver Stress Signal Calibration Setup 2 N5990A Test Automation SW for PCIe PCIe Base Specification 4.0 requires a CEM connector to part of the test channel! PCIe 4.0 Base Spec requires a CEM connector to be part of the test channel Channel calibration with preset selection to get as close to target eye height and eye width as possible. J-BERT M8020A s internal ISI can be used to calibrate channel. Preset is selected on optimal RX eye area DM-SI and CM-SI are calibrated through the channel Compliance eye calibration is done by adjusting DM-SI, SJ or V diff If SJ was changed from 100 mui during the compliance eye calibration, the SJ portion >100 mui is applied as a secondary SJ tone @ 210 MHz during the RX test. This allows to follow the jitter tolerance compliance curve PCIe 4.0 16GT/s RX Testing Page 45
PCIe 4.0 RX Stress Signal Calibration Channel Calibration Channel closest to 28dB loss @ 8GHz is chosen initially Optimal preset is chosen for channel setting based on RX eye area after reference RX Insertion loss of channel is varied to get close to target eye height and target eye width Reference RX eq is optimized by analysis SW for each channel + preset combination Too many steps for manual procedure PCIe 4.0 16GT/s RX Testing Page 46
PCIe 4.0 RX Stress Signal Calibration Compliance Eye Calibration Channel determined by the channel calibration is applied DM-SI, SJ and V diff are adjusted to find correct combination for a compliant eye of the reference RX PCIe 4.0 16GT/s RX Testing Page 47
Conclusions 1. Gen4 is at the 0.7 release version and work is underway on v0.9. Most of the hard work has been done but it is important for the SIG to ensure the details are captured. 2. Official statement on scope BW required for TX and RX calibration is in the v0.9 spec. The official requirement is a minimum scope bandwidth of 25GHz due to the need to match edge BW of JBERT during calibration. This also is required for TX testing. 3. PCIe 4.0 Reference Clock Phase Jitter requirements have expanded significantly since PCIe 3.0. 4. Official U.2 fixtures from the PCISIG are now available (SFF-8639) 5. Receiver Calibration Channel is to have a CEM connector EIEOS goes to 16 1 s and 16 0s at v0.7. 6. PCIe 4.0 informal FYI testing (CEM) will begin at the April workshop in Milpitas. 7. Tools for full PCIe 4.0 TX and RX BASE testing are available today. Keysight has the only demonstrated Gen4 capable, protocol aware BERT for RX testing at 16GT/s. PCIe 4.0 16GT/s TX/RX Testing Page 48
PCI Express 4.0 Keysight Total Solution Physical layer interconnect design ADS design software 86100D DCA-J/TDR E5071C ENA option TDR Industry s lowest scope noise floor/sensitivity and trigger jitter PCIe 4.0 16GT/s TX/RX Testing Page 49
PCI Express 4.0 Keysight Total Solution Physical layer interconnect design Physical layertransmitter test ADS design software 90000 X, Z-Series oscilloscope 86100D DCA-J/TDR N5393F PCI Express electrical compliance software E5071C ENA option TDR 86100CU-400 PLL and Jitter Spectrum Measurement SW Industry s lowest scope noise floor/sensitivity and trigger jitter DSA-X Series & Q Series Real-Time Oscilloscopes PCIe 4.0 16GT/s TX/RX Testing Page 50
PCI Express 4.0 Keysight Total Solution Physical layer interconnect design Physical layertransmitter test Physical layerreceiver test ADS design software 90000 X, Z-Series oscilloscope M8020A J-BERT High Perfformance BERT 86100D DCA-J/TDR N5393F PCI Express electrical compliance software N5990A automated compliance and device characterization test software E5071C ENA option TDR 86100CU-400 PLL and Jitter Spectrum Measurement SW Industry s lowest scope noise floor/sensitivity and trigger jitter DSA-X Series & Q Series Real-Time Oscilloscopes Automated compliance software accurate, efficient and consistent PCIe 4.0 16GT/s TX/RX Testing Page 51
For more Information You will find more information on PCI Express and Keysight test solutions at: www.pci-sig.com www.keysight.com/find/pciexpress PCI-SIG Website, Specification, S/W Tools, Keysight Test Procedure Keysight tools to help you succeed with your PCI Express design such as the N5393F Compliance application. http://electronicdesign.com/keysight/digital-design-and-test-webcast-series Keysight digital webcast registration page. www.keysight.com/find/pcie_receiver_test PCIe 4.0 Rx Test Detailed Information Rick Eads Principal PCI Express Planner rick_eads@keysight.com PCIe 4.0 16GT/s RX Testing Page 52
Invitation to Join PCI-SIG Group on Linked-In* Must be employed by a member company of the PCI-SIG to join. PCIe 4.0 16GT/s TX/RX Testing Page 53