CCS-Model-Introduction

CCSModelIntroductionHuangRuifengTianHongshengOutline•Background•CCSTimingModel•CCSNoiseModel•CCSPowerModel•CCSModelSettinginNanoTimeBackground•LimitationofNLDMModel–Non-LinearDelayModelDriverModel:ATransitionTimeReceiverModel:ACapLoadBackground•LimitationofNLDMModel–Atprocessgeometriesof90-nmandbelow,manyneweffectscannotbeaccuratelymodeledusingthisapproach.•HighImpedanceInterconnect•Millereffect•DynamicIR-drop•Multi-voltage,andDynamicVoltageandFrequencyScaling(DVFS)•Driverweakening•Temperatureinversion•Increasingvariations•Synopsys:CCS–CompositeCurrentSourceModel•Cadence:ECSM–EffectiveCurrentSourceModel•ThesemodelsaremoreaccuratethanNLDMforadvancedprocesstechnologies(90nm)BackgroundCCSModel•Synopsys:CCS–CompositeCurrentSourceModel–Handlehigh-impedancenetsandothernonmonotonicbehavior.–Current-based,itenablesbothtemperatureandvoltagescalingofallcellbehavior.–Canbeappliedtotiming,noiseandpower.–Moresophisticatedthanconventionalinterpolation,thusimprovingaccuracy.•CCSTimingModel•CCSNoiseModel•CCSPowerModelCCSModelCCSTimingModelCCSTimingModel•CharacterizationforCCSTimingICoutforDriverModel,NotTransitionTimeIinpforReceiverModel,OneTwoCapLoadCCSTimingModel•CCSDriverModelHowtosegmenttheICout?•CCSTimingDriverModelRequirements–Simulationautostopandfulltransition•Outputpinvoltageiswithin1mVofrail•Integratesthecurrentvectorshouldreachwithin5%oftherail–Thecurrentvectorshouldnotcontainzerocurrentvalues.–Currentpolarity•Fortheoutput_current_risegroup,valuesmustbepositive.•Fortheoutput_current_fallgroup,valuesmustbenegative.–Peakcurrent•FirstandlastcurrentvectorsholdnotbethepeakofthecurrentCCSTimingModel•CCSTimingDriverModelRequirements–Currentvectorvariables•Variable_1:input_net_transition•Variable_2:total_output_net_capacitance•Variable_3:time–Starttimeandreference_timeCCSTimingModelIdenticalreference_timeforthesameinput_net_transition.•CCSTimingDriverModelRequirements–ModifyfirstcurrentvaluetoallowVoltagetoexactlymatchatfirstsavedtimepointCCSTimingModelintegratedvoltagematchesactual•CCSTimingDriverModelRequirements–Integratethecurrentwaveformandcheckthatthevoltageiswithin5%ofthefinalrailvoltage.CCSTimingModel•CCSReceiverModel–Pin-based–Arc-based–TwocapvaluestomodelthevariationofcellinputCCSTimingModel•CCSReceiverModel–Whyweuse2capstomodel?CCSTimingModelC_total=0.40+0.45=0.85fC_total=0.54+0.35=0.89f•CCSReceiverModel–HowtocharacterizeC1andC2?CCSTimingModelS1S2•CCSReceiverModel–Arc-based–Pin-basedCCSTimingModel•CCSNofaChannel-ConnectedBlock(CCB)–CCSnoiseDCcurrenttableCCSNoiseModeldc_current(ccsn_dc_template){index_1(“input_voltage,-VDD~2*VDD”);index_2(“output_voltage”,-VDD~2*VDD”);values(......);}•CCSNofaChannel-ConnectedBlock(CCB)–CCSNstageoutputvoltagetables•TheaccuracyofaCCSnoisemodelisnotverysensitivetotheexactvaluesofinputslewandloadcapacitancethatyouchoose.CCSNoiseModel~10%~30%~50%~70%~90%output_voltage_fall(){vector(ccsn_vout_template){index_1(“input_net_transition”);index_2(“total_output_net_capacitance”);index_3(“time”);values(......);}•CCSNofaChannel-ConnectedBlock(CCB)–CCSNpropagationtablesCCSNoiseModelpropagated_noise_high(){vector(ccsn_prop_template){index_1(“input_noise_height”);index_2(“input_noise_width”);index_3(“total_output_net_capacitance”);index_4(“time”);values(......);}peak80%peak50%peak•CCSNofaChannel-ConnectedBlock(CCB)–CCSNpropagationtables•Howtofindappropriateinputnoiseheightandwidthvalues?CCSNoiseModel•CCSNofaChannel-ConnectedBlock(CCB)–MillercapacitancecharacterizationCCSNoiseModelQ=△Vout*CM=Cin,1*△Vin,1=Cin,2*△Vin,2•CCSNofMultipleChannel-ConnectedBlock(CCB)–Arc-basedCCS•usethearc-basedCCSwheneverpossiblebecauseitallowsmoredetailednoiseanalysis–Pin-basedCCS•Arcdescribesa3-ormore-stagepath.•Existenceofasensitizable3-ormore-stagepathbetweenthetwoarcpins.•Electricalreconvergencesomewherebetweenthethearcfromandtopins.•NanoTimesupports(only)pin-basedCCSnoisemodels.CCSNoiseModel•tobecontinueCCSPowerModel•extract_model-nameDMA-library_elements{nldmccs_timing}•setsim_use_snps_predrivertrue•setmodel_ccs_characterization_driver_typesnps_predriverCCSTimingSettinginNanoTime•setmodel_ccs_measured_delay_tolerance0.02(default)•setmodel_ccs_waveform_segment_tolerance0.005(default)CCSTimineSettinginNanoTime•setsi_enable_noise_analysistrue•extract_model-nameDMA-library_elements{nldmccs_timingccs_noise}•setsim_file_cleanup0-6(0savealland6savefailedsimulations)•setlib_current_unit1uA•extract_model-hspice_timingCCSNoiseSettinginNanoTime谢谢！