Keysight Technologies Inc. · NYSE:KEYS
Keysight Technologies — Technical Presentation (2025)
50 slides from Keysight Technologies's technical presentation, Feb 26, 2025
Slide 1: Systematic and accurate crosstalk analysis
Systematic and accurate crosstalk analysis Shaishav Pandya / Keysight Circuit Sim HSD Software Architect SK Choi / Keysight Circuit Sim HSD Product Owner 1
Slide 2: Agenda
Agenda • Quick recap on Memory systems • Signal Integrity Impairments – Aspiration Vs Reality • Crosstalk – NEXT & FEXT • Smart and accurate crosstalk analysis • Case study • Final Takeaways Aggressor Victim NEXT FEXT Take the guesswork out of crosstalk mitigation in memory systems! 2
Slide 3: Quick recap on Memory buses
Quick recap on Memory buses 1. Point to Point connection: • Data bus – Single ended signals • Data strobe – Differential signals 2. Fly-by connection • Command address bus • Command/Address signal – single ended signals • Clock signal - Differential • Control signal • Clock, address and control sign
Slide 4: Typical Memory System
Typical Memory System Write cycle DQ0 During a Write cycle the During a read cycle, the controller is transmitting Memory Memory is transmitting Controller data, and the memory is data, and the Controller is the receiver of the data DQ63 the receiver of the data Read cycle Equivalent Setup in ADS Me
Slide 5: Typical SI Measurements for a Write Cycle
Typical SI Measurements for a Write Cycle DQ6 DQ7 DQ8 DQ7 DQ8 Signal integrity metrics Signal integrity metrics • Eye Diagrams at Controller Write cycle • Eye Diagrams at receiver Signal integrity metrics • Crosstalk • Frequency dependent losses (IL,RL) 5
Slide 6: Aspiration Vs Reality
Aspiration Vs Reality DQ6 DQ7 DQ8 Write cycle Expect the eye to be open! Expect Eye to look completely Expect Eye to be open at the open at the Transmitter Receiver 6
Slide 7: Aspiration vs Reality
Aspiration vs Reality DQ6 DQ7 DQ8 Write cycle Simulated eye is almost closed! Amplitude noise Eye is almost closed on Rx 7
Slide 8: Eye Shows Reflection, Crosstalk, and Considerable Loss
Eye Shows Reflection, Crosstalk, and Considerable Loss Received eye at RxData7 without equalization 1. Signs of reflection 2. Signs of crosstalk 3. Signs of frequency- dependent loss Rule Number 9: Anticipate Before You Measure or Simulate* *E. Bogatin, "Bogatin's 20 Rules for Engineers," Signal Int
Slide 9: Causes and Effects of Reflection, Crosstalk, and Considerable Loss
Causes and Effects of Reflection, Crosstalk, and Considerable Loss Degradation of both the eye amplitude and shape. 1. Signs of reflection Cause: Impedance discontinuity 2. Signs of crosstalk ? Rise/Fall time degradation which 3. Signs of frequency- reduces both timing margin and dependent loss volt
Slide 10: Eye Shows Reflection, Crosstalk, and Considerable Loss
Eye Shows Reflection, Crosstalk, and Considerable Loss Received eye at RxData7 without equalization 2. Signs of crosstalk 10
Slide 11: Eyes Are Open When the Channel Are 10 Trace-width Apart
Eyes Are Open When the Channel Are 10 Trace-width Apart TxData6 RxData6 Eyes are open when Spacing is 10x trace width the channels are 10 trace width apart. TxData7 RxData7 Spacing is 10x trace width There is no crosstalk when the channels are far from each other. TxData8 RxData8 11
Slide 12: Expect Near-end Crosstalk When Channels Are Close
Expect Near-end Crosstalk When Channels Are Close 1 TxData6 RxData6 0 Spacing: 1x trace width (1W) TxData7 RxData7 Terminated Terminated 1 0 Crosstalk voltage and signal Active Current loop: clock-wise arrive at the same time TxData6 RxData6 TxData7 RxData7 Terminated Induced current loop: counter c
Slide 13: Near-end Crosstalk Signatures in the Waveform
Near-end Crosstalk Signatures in the Waveform Crosstalk voltage signature in Channels are close and TxData7’s waveforms they are all active TxData6 RxData6 Spacing: 1 trace width TxData7 RxData7 Spacing: 1 trace width TxData8 RxData8 13
Slide 14: Near-end Crosstalk Signatures in the Eye
Near-end Crosstalk Signatures in the Eye Crosstalk voltage signature in Channels are close and TxData7’s eye they are all active TxData6 RxData6 Spacing: 1 trace width TxData7 RxData7 Spacing: 1 trace width TxData8 RxData8 14
Slide 15: Near-end Crosstalk Signatures in the Eye
Near-end Crosstalk Signatures in the Eye Crosstalk voltage signature in Channels are close and TxData7’s eye they are all active TxData6 RxData6 Spacing: 1 trace width TxData7 RxData7 Spacing: 1 trace width TxData8 RxData8 Near-end crosstalk appears as thickening of eye levels. 15
Slide 16: Expect Far-end Crosstalk When Channels Are Close
Expect Far-end Crosstalk When Channels Are Close 1 TxData6 RxData6 0 Spacing: 1x trace width (1W) TxData7 RxData7 Terminated Terminated 1 0 Crosstalk voltage and signal Active Current loop: clock-wise arrive at the same time TxData6 RxData6 TxData7 RxData7 Terminated Induced current loop: counter cl
Slide 17: Far-end Crosstalk Signatures in the Waveform
Far-end Crosstalk Signatures in the Waveform Channels are close and Crosstalk voltage signature in they are all active RxData7’s waveforms TxData6 RxData6 Spacing: 1 trace width TxData7 RxData7 Spacing: 1 trace width TxData8 RxData8 17
Slide 18: Far-end Crosstalk Signatures in the Eye
Far-end Crosstalk Signatures in the Eye Channels are close and Crosstalk voltage signature in they are all active RxData7’s eye TxData6 RxData6 Spacing: 1 trace width TxData7 RxData7 Spacing: 1 trace width TxData8 RxData8 18
Slide 19: Far-end Crosstalk Signatures in the Eye
Far-end Crosstalk Signatures in the Eye Channels are close and Crosstalk voltage signature in they are all active RxData7’s eye TxData6 RxData6 Spacing: 1 trace width TxData7 RxData7 Spacing: 1 trace width TxData8 RxData8 Far-end crosstalk shows up as peaks in the transitions 19
Slide 20: Far-end Crosstalk Signature Looks Different
Far-end Crosstalk Signature Looks Different Received eye at RxData7 Received eye example without equalization 20
Slide 21: Previous Channels Have the Same Lengths
Previous Channels Have the Same Lengths 1 TxData6 RxData6 0 Spacing: 1x trace width (1W) TxData7 RxData7 Terminated Terminated 21
Slide 22: Change the Channel Length to Offset the Timing of Crosstalk
Change the Channel Length to Offset the Timing of Crosstalk 1 TxData6 RxData6 0 Spacing: 1x trace width (1W) TxData7 RxData7 Terminated Terminated 1 0 Crosstalk voltage and signal Active Current loop: clock-wise arrive at different times TxData6 RxData6 Crosstalk voltage arrives at different times w
Slide 23: Xtalk and Length Variation Cause Vertical Eye Collapse
Xtalk and Length Variation Cause Vertical Eye Collapse Received eye example Received eye example With same channel lengths With different channel lengths Crosstalk with channel length variations can cause vertical eye collapse! 23
Slide 24: Xtalk and Length Variation Cause Vertical Eye Collapse
Xtalk and Length Variation Cause Vertical Eye Collapse 3 nearby channels separated by one trace- width and with all different lengths TxData6 RxData6 Eye 6 TxData7 RxData7 TxData8 RxData8 Eye 7 Crosstalk from multiple channels with multiple channel length variations can cause vertical eye collapse.
Slide 25: Impact of NEXT and FEXT– Which One Has More Impact?
Impact of NEXT and FEXT– Which One Has More Impact? EYE Q: Given same level of NEXT and FEXT, which A: It depends. one degrades the eye more? 25
Slide 26: Impact of NEXT and FEXT– Tx on the Same Side
Impact of NEXT and FEXT– Tx on the Same Side EYE FEXT has more impact Tx Rx 26
Slide 27: Impact of NEXT and FEXT – Tx on the Opposite Side
Impact of NEXT and FEXT – Tx on the Opposite Side EYE NEXT has more impact Rx Tx 27
Slide 28: Near-end Crosstalk Signatures in the Eye
Near-end Crosstalk Signatures in the Eye Channels are close and Crosstalk voltage signature in they are all active RxData7’s waveforms TxData6 RxData6 Spacing: 1 trace width TxData7 RxData7 Spacing: 1 trace width TxData8 RxData8 28
Slide 29: Summary of Crosstalk on Eye Diagram
Summary of Crosstalk on Eye Diagram Eye thickens at Tx due Peaks in eye transitions to NEXT at Rx due to FEXT 29
Slide 30: Fix Near-end Eye Closure by Increasing Channel Spacing
Fix Near-end Eye Closure by Increasing Channel Spacing Signs in the Eye Root cause Fixes • Thickening of the eye Crosstalk from adjacent • Move channels farther apart levels channels • Keep TXs on side same* *Does not eliminate near-end crosstalk, but shifts focus to far-end 30
Slide 31: Fix Far-end Eye Closure by Increasing Channel Spacing
Fix Far-end Eye Closure by Increasing Channel Spacing Signs in the Eye Root cause Fixes • Move channels farther apart • Peaks in eye transitions Crosstalk from adjacent • Maintain the same channel • Lumped peaks in the eyes channels length 31
Slide 32: Improve Your Signal Integrity by Removing the Root Cause
Improve Your Signal Integrity by Removing the Root Cause Signs in the Eye Root cause Fixes • Deformed eye shape • Locate impedance discontinuity and Impedance discontinuities • Reduced eye amplitude match it with channel impedance • Peaks in eye transitions (FEXT) • Move channels farther apart Cross
Slide 33: Agenda
Agenda • Quick recap on Memory systems • Signal Integrity Impairments – Aspiration Vs Reality • Crosstalk – NEXT & FEXT • Smart and accurate crosstalk analysis • Case study • Final Takeaways 33
Slide 34: Old Way for Crosstalk Analysis
Old Way for Crosstalk Analysis Crosstalk simulation setup Very complicated setup for entire 8 Bytes Transmission line Static Transmission line termination values Transmission line Transmission line Too much hassle!! Need a more systematic way of setting up crosstalk simulations in Memory systems! 34
Slide 35: Crosstalk Analyzer
Crosstalk Analyzer A new smarter way of analyzing crosstalk effect Termination Streamlined provided by setup and IBIS file workflow! HTML Report Generation! 35
Slide 36: Crosstalk Analysis Setup
Crosstalk Analysis Setup In Memory Designer Aggressor modes: • Continuous • Single bit • Step: Rise and Fall Memory Probe • Choose the list of aggressors to be includes in the analysis • Choose the list of test points where the measurements are made • Setup crosstalk analysis report generation (HTML
Slide 37: Comparison of different modes
Comparison of different modes Aggressor mode Continuous bit aggressor Single bit aggressor Rise step mode aggressor Fall step mode aggressor 37
Slide 38: Crosstalk Analyzer Simulation Process
Crosstalk Analyzer Simulation Process … Step 3 1 2 Aggressor Victim (Silent) DQ6 Victim Victim Victim (Silent) Aggressor (Silent) DQ7 Victim Victim (Silent) Aggressor Victim (Silent) DQ8 Victim NEXT … FEXT 38
Slide 39: HTML Report generation
HTML Report generation Pass/Fail Consolidated report with pass/fail test results for every victim test point! 39
Slide 40: Viewing results from Crosstalk Analyzer
Viewing results from Crosstalk Analyzer 40
Slide 41: Test Case: Byte7 from FPGA to DRAM
Test Case: Byte7 (DQ48~DQ55, DQS6, and DM6) from FPGA to DRAM Xilinx Versal Platform Versal Physical Board (EM modeled with SIPro) Memory Designer Schematic Representation Versal: IBIS models S-parameter DDR Speed Grade: 3200Mbps Micron: IBIS models 41
Slide 42: Eye Diagram Validation at DRAM Input
Eye Diagram Validation at DRAM Input Speed Grade: 3200Mbps Memory Probe • Access any nodes • Measure SI metrics Eye measure points Eye Height = 447~513 mV Eye Width = 245.3 ~ 269 psec 42
Slide 43: Crosstalk Analysis Results By Test Points
Crosstalk Analysis Results By Test Points Using ADS Data Display FPGA DIMM Input DIMM Output DRAM Input DRAM Die 43
Slide 44: Crosstalk Analysis Results By Test Points
Aggressor Crosstalk Analysis Results By Test Points 100mV+ Detail look 50~100mV 10~50mV • Aggressor: DM6 signal >10mV • Aggressor mode: Step Rise • Victim model: High State DQ48 DQ49 DQ50 DQ51 DQ52 DQ53 DQ54 DQ55 DM6 DQS6 FPGA(C0) 123.5 23.30 11.75 7.971 11.43 16.14 16.33 5.329 909.5 15.69 DIMM inpu
Slide 45: Crosstalk Analysis Results By Test Points
Crosstalk Analysis Results By Test Points A few interesting findings • With DM6 to be the aggressor, • DQ48 has highest crosstalk to DM signals DQ48 123.5 mVpp DQ49 23.3 mVpp DQ49 19.2 mVpp DQ48 • DQ49 is nearer than DQ48 but higher crosstalk 108.3 mVpp • DQS has relatively small crosstalk • Less th
Slide 46: Victim's State Changes Crosstalk Voltage Level
Victim’s State Changes Crosstalk Voltage Level Accurate loading/termination Impedance • Aggressor: DM6 signal • If the victim (DQ48) is in high state, the peak-to-peak crosstalk voltage is 123.5mV 9.8mV difference • If the victim (DQ48) is in low state, the peak-to-peak crosstalk voltage is 133.2mV
Slide 47: Final Takeaways
Final Takeaways • Crosstalk is a critical test item in HSD (High-Speed Digital) design. • Single-ended signals are more significantly affected by crosstalk than differential signals, making thorough verification essential. • Keysight ADS's Crosstalk Analyzer feature enables a more systematic approac
Slide 48: Thank you
Thank you
Slide 49: References
References • Introduction to IBIS (I/O Buffer Info Specification) Modeling 49
Slide 50: vy KEYSIGHT
vy KEYSIGHT [ \
Keysight Technologies — Technical Presentation (2025) on 6ix