docPredicting the Performance of Sort Tooling to 40 Gb/s
download 
Report Transcription
Predicting the Performance of Sort Tooling to 40 Gb/s
Predicting the Performance of Sort Tooling to 40 Gb/s Brett Grossman Signal Integrity Staff Engineer Intel Corporation Hillsboro, OR [email protected] Bryan Boots Application Engineer Ansoft Corporation Boulder, CO [email protected] Southwest Test Workshop - June 1, 2003 1 Outline • • • • • • • Background Problem Statement Model creation process Model correlation Sensitivity analysis Summary Acknowledgements Southwest Test Workshop - June 1, 2003 2 Telecom Datarate Trend 45 OC-768 40 IO Data Rate (Gbps) 35 30 25 20 •Sonet •Sonet==Synchronous SynchronousOptical OpticalNetwork, Network, •OC Carrier •OC==Optical OpticalCarrier •OC-xxx •OC-xxx==Data Datarate rate➨➨‘xxx’ ‘xxx’* *51.84 51.84Mbps Mbps 15 OC-192 10 5 0 Firewire USB OC-48 Serial ATA Infiniband Serial ATA USB2 Year Time Southwest Test Workshop - June 1, 2003 3 What is 40 Gb/s? • PMD devices – PMD – Physical Media Dependent • 40 Gb/s devices characterized by: – Very small risetime (~10ps) – Very small amplitude (~10mV) – Very small die size (~1.0 mm) Southwest Test Workshop - June 1, 2003 4 40 Gb/s - another view Equivalent to transmitting more than 7 CD-ROM’s of data every second Southwest Test Workshop - June 1, 2003 5 40 Gb/s – yet, another view Southwest Test Workshop - June 1, 2003 6 Probe Technology Continuum 10% of Probe Technologies # of technologies 90% of Probe Technologies DC 50 GHz 110 GHz • > 90% of probe technologies currently address 10% of this spectrum • Illustration based on response to RFI submitted to 35 probe technology companies Southwest Test Workshop - June 1, 2003 7 Probe Evaluation • Test vehicle designed – For the typical battery of tests • Lifetime • Cres • … • But, high frequency performance was critical Top view of evaluation vehicle How could we evaluate the impact of design trade-offs on high frequency performance? Southwest Test Workshop - June 1, 2003 8 Problem Statement • Need to quantify performance tradeoffs – Performance = ƒ(suppliers experience) • Desired robust process where some iterations could be ‘virtual’ • Current tools/methods had not demonstrated capability to support this complexity Southwest Test Workshop - June 1, 2003 9 Model Topology Instrument coax Center line coax Connector Mechanical Assembly Flex PCB Probecard cross section Probe DUT • Complex signal path topology – Multiple substrates (cable, PCB, PI flex) – Multiple guide structures (Coax, CPW, microstrip) • Physically small, Electrically large features Southwest Test Workshop - June 1, 2003 10 Model Creation • Two high speed channels selected • Signal paths segmented into 3 logical blocks Layout for flex component 1. Flex block 2. Coax block 3. Transition block High speed channels on flex Southwest Test Workshop - June 1, 2003 11 Model Creation (flex block) GDS • Need to translate directly from physical layout HFSS GDS Boolean HFSS GDS LinkCAD HFSS GDS Ansoft Designer AnsoftLinks HFSS Other names and brands may be claimed as the property of others Southwest Test Workshop - June 1, 2003 12 Model Creation (coax block) • Model for connector and coax built from coax structure and data sheet parameters • Model parameters (blue) fit to measurements (red) Coax model Southwest Test Workshop - June 1, 2003 13 Model Creation (transition block) • The transition elements were created manually in HFSS • The individual elements are assembled for simulation Merged model Southwest Test Workshop - June 1, 2003 14 Model Correlation Thru measurement (red) to model (blue) correlation shown from 0.05-40.05 GHz Channel #1 Channel #2 Southwest Test Workshop - June 1, 2003 15 Sensitivity Analysis – Microstrip vs. CPW length Southwest Test Workshop - June 1, 2003 16 Sensitivity Analysis • Impact on insertion loss Offset Southwest Test Workshop - June 1, 2003 17 Sensitivity Analysis Southwest Test Workshop - June 1, 2003 18 Summary • HFSS Learning curve, creating a robust modeling ‘process’ • Ended up successfully modeling the path based on physical geometry • Can continue to use this model for what-if type analysis to assess design trade-offs Southwest Test Workshop - June 1, 2003 19 Acknowledgements • The authors would like to thank Marc Berube for his continued support of this work. • We would also like to thank Cascade Microtech for their contributions and open discussions. Southwest Test Workshop - June 1, 2003 20 References • • • • • Ansoft – www.ansoft.com Cascade Microtech – www.cmicro.com Boolean - www.xs4all.nl/~kholwerd/bool.html LinkCAD – www.linkcad.com Intel – www.intel.com , developer.intel.com Southwest Test Workshop - June 1, 2003 21
