Flicker Noise Extraction for Scalable MOS …

1 FlickerNoise Extraction forScalableMOS simulation ModelsDr. ThomasGneitingAdMOS Flickernoise equationsin modern MOSmodels Noise simulationin differentsimulators Stepsinparameter Extraction usinga PSPexample Problems tobe solved SummaryPage3 Common FlickerNoiseModels10010110210310410510-1 710-1810-1910-2010-2110-2210-2310-24f [Hz]Sid[A /Hz]FrequencyexponentStart ofthermal noise10010110210310410510-1710-1810-1910 -2010-2110-2210-2310-24f [Hz]Sid[A /Hz]DC biasdependency+ DimensiondependencyThe most MOS simulation models provide the following 2 principaldegrees of freedom to adjust the Flicker FlickerNoiseEFeffOXAFDSidfLCIKFfS 2)( *282222141482210*22102102log10*2)(NNNNOI CNNOIBNOIAfLINTNOILWLITkNNNOICNNNOIBNNNO IAfALINTNOILCIqTkfSlllEFeffeffCLMDSBlOlO lOEFbulkeffoxeDSeffBid 362''104)(,)( EFeffeffDStmwiwimitliwimitliidfLWIVNOIAf SSSSSfSNOIMOD=1,4,5 NOIMOD=2,3,6 Vgs>Vth+ :Vgs<Vth+.

2 Classic SPICE modelBSIM3 modelPage5 BSIM4 FlickerNoiseEFeffOXAFDSidfLCIKFS 2 2*2102222**1022,10*22log10*2)(NNNNOICNNO IBNOIAfLINTNOILWLITkNNNOICNNNOIBNNNNNOIA fALINTNOILCIqTkfSlllEFeffeffCLMDSBlOlOlO EFbulkeffoxeDSeffBinvid 102*2,,,,,10)(,)()()()()( NfLWITkNOIAfSfSfSfSfSfSEFeffeffDSBsubVti dsubVtidinvidsubVtidinvididFNOIMOD=0 FNOIMOD=1 Inversion:SubVth: classic SPICE modelPage6 PSP FlickerNoise NNNNFCNFBNNNNNNFCNNFBNFANGCfIqfSmmmvsato xDSTfl **2**222/2/ln)( EEENENEEENENEEENENLWLWNFCLWNVCLWLWNFBLWN FBLWLWNFALWNFA Localmodel:Scalingequation:Fixed frequency exponentof 1!

3 Page7 Noisein ADSdefine noise_circ(D G OUT)Flicker_Noise:MAIN D1 G 0 0\L=10u W=10uMult=1;; sensethe Noise currentR:RdummyD D1 R=1mNoise=no;;convert Noise current intoanidentical voltageZ_Port:v D1 D OUT 0 Z[2,1]=1R:RauxOUT 0 R=1kNoise=no;endnoise_circPage8 NoiseinSpice, noise_circ1 2 5 7 MAIN 9 6 0 0 Flicker_Noise+ L=10u W=10u NF=1L2 6 2 1000000C2 5 6 1000000H1 8 0 V1 1R1 7 8 1e-4V1 1 9 Flickernoise equationsin modern MOSmodels Noise simulationin differentsimulators Stepsinparameter Extraction usinga PSPexample Measurement(see paper fromIHP, Falk Kornd rfer)

4 Determination ofthe frequency exponent Determination ofbias dependency Extractionofparameters for Scalable Noise models shown withanexample using thePSPmodel Problems tobe solved Summary*)Basedonthe workofKnoblinger,Grabinski,Sischka*)Page 10 FrequencyExponent (1)10010110210310410510-1710-1810-1910-2 010-2110-2210-2310-24f [Hz]Sid[A /Hz]EFIf the model provides afrequency exponent, it can bederived from the slope of themeasured curves in logarithmicrepresentation:(Example for SPICE model, the othermodels behave similar) xmcyfEFconfSfconfSfLCIKFfSidEFidEFeffOXA FDSid loglog)(log)()(2 Page11 FrequencyExponent (2)In the real world: EF must fitseveral deviceswith diff.

5 L, Wsimultaneously Take intoaccount onlyparts of thecurves due Biasdependency(1)10010110210310410510-17 10-1810-1910-2010-2110-2210-2310-24f [Hz]Sid[A /Hz]Sid[A /Hz]0123410-1710-1810-1910-2010-21Vg[V]O nce the slope is given, the Noise values at 1Hz can be determined byextrapolating the measured curve to 2)(=1 at 1Hz(Example for SPICE model, the othermodels behave similar)Page13DC Biasdependency(2)3 Steps: Select range ofcurves todetermine slope Make curvesslopeless Apply linearfitting to theselected range ofthe curve anddetermine cross-point at 1 HzPage13DC Biasdependency(2)3 Steps: Select range ofcurves todetermine slope Make curvesslopeless Apply linearfitting to theselected range ofthe curve anddetermine cross-point at 1 HzPage13DC Biasdependency(2)3 Steps.

6 Select range ofcurves todetermine slope Make curvesslopeless Apply linearfitting to theselected range ofthe curve anddetermine cross-point at 1 HzPage13DC Biasdependency(2)3 Steps: Select range ofcurves todetermine slope Make curvesslopeless Apply linearfitting to theselected range ofthe curve anddetermine cross-point at 1 HzPage14 ParameterExtraction PSP (1)Sid(f=1Hz)@ diff Vg,Vd, L, WSolverSfl(1Hz)=f(Parameter,Dimensions,D C Bias)Iterative Solutionusing a modifiedLevenberg-Marquardt algorithmNoise parametersPSP:PSP1020 , W, MULTVG, VDPSP:NFALW, NFBLW,NFCLWPage15 ParameterExtraction PSP (2) withdefault valuesof NFALW,NFBLW, inca.

7 1s andrepeatsimulationExample parameterextraction process:Page15 ParameterExtraction PSP (2) withdefault valuesof NFALW,NFBLW, inca. 1s andrepeatsimulationExample parameterextraction process:Page15 ParameterExtraction PSP (2) withdefault valuesof NFALW,NFBLW, inca. 1s andrepeatsimulationExample parameterextraction process:Page16 ParameterExtraction PSP (3)Final resultfor tobe solvedThe scalability of the simulation models isstill not good PSP scaling equation:does not allow to modify parametersindependently for L and W.

8 Actually, onlythe product L W is taken into Fittingfor smalldevice cannot Together with the measurement of Flicker Noise ( presentedby FalkKornd rfer, IHP), the shown Extraction strategy isthe basis of a complete noisemodelingmethod. The very effective simultaneous Extraction of Flicker noiseparameters for different devices with different dimensions isthe key improvement of this tool. The shown methodology for PSP was implemented forcommon MOS models (BSIM3, BSIM4) and can be easilyextended to other models like HiSIM2 etc.

9 The co-operation between IHP and AdMOS resulted in acommercial available Flicker NoiseModelingTool. Fordetails, please Products Flicker Noise System