Home

user manual

1. Y n X n I The convergence check tightens as the value of this parameter is increased ntoltype Integer 3 this option specifies the tolerance check adopted at each time point of the analysis to control if the Newton method has reached convergence Checks are done both on the variation of variables computed by the Newton method and on the entries of the right hand side vector Define as dX n the variation of the n electrical variable as reltol and abstol the relative and absolute maximum allowed variations respectively The n th variable has converged if dX n lt reltol Y n abstol The ntoltype parameter defines three different ways to compute Y n number as 1 2 and 3 Their meanings are listed in the following 1 global Y n is defined as the maximum over the time interval till the current time point of the maximum value assumed by the X n electrical variables where n varies from 1 to N being N the number of circuit variables Note that in this case Y n is unique value adopted for all the dX n variable variations 2 local Y n is defined as the maximum absolute value of X n in the time interval till the current time point Note that in this case Y n is different for each electrical variable 3 point local Y n X n I The convergence check tightens as ntoltype is increased Default value of this parameter
2. degree of annotation Values from 1 to 7 turn on display of generic information about the analysis run Values above 7 are bitwise and turn on specific printouts 0x08 printout the state transition matrix start Start sweep limit stop Stop sweep limit center Center of sweep span Sweep limit span step Step size linear sweep lin Number of points linear sweep dec Points per decade log Number of points log sweep sweeptype Integer 0 type of frequency sweep Possible ones are 0 absolute sweep 1 relative sweep 2 absolute offset sweep 3 relative offset sweep When absolute sweep is chosen the frequency is swept in the interval defined exclusively 4 8 5 4 8 Advanced time domain analyses 135 by the sweep When the relative sweep is chosen the frequency is swept in the interval specified by the sweep with respect to the fundamental frequency corresponding to the working period determined by the large signal steady state analysis In other words the frequency of the sweep adds to the circuit large signal fundamental When the absolute offset sweep is chosen the frequency is swept as described in the absolute sweep mode but the results are referred to the frequency of the offset sweep i e the output frequency does not consider the fundamental and harmonics of the circuit When the relative offset sweep is chosen the frequency is swept as described in the relative sweep mode but the
3. MODEL DEFINITION model ModelName tline lt parameter value gt MODEL PARAMETERS f Reference frequency vel Propagation velocity normalized to c r Series resistance per meter at fc g Shunt conductance per meter corner Skin depth corner frequency alphac Conductor loss at fc alphad Dielectric loss dcr DC series resistance per meter qc Conductor loss quality factor qd Dielectric loss quality factor fc Conductor loss measurement frequency fd Dielectric loss measurement frequency Voltage controlled switch The voltage controlled switch is tied between terminals n1 and n2 and it is driven by the voltage across the n3 and n4 terminals The switch resistance varies from the off value switch is off to the on one switch is on This variation is linear with respect to the controlling voltage in the voff von range The noise model refers to the varying output resistance of the voltage controlled switch Input resistance in noiseless ELEMENT DEFINITION Name nl n2 sense sense ModelName lt parameter value gt ELEMENT PARAMETERS noisetype Integer 5 types of generated noise possible choices are 0x0 noiseless voltage switch Ox1 thermal noise 0x4 flicker noise The value of this parameter can be the result of the bitwise or of types of noise sources MODEL DEFINITION model ModelName vswitch lt parameter value gt MODEL PARAMETERS ron Real 1 0e 3 on resistance The switch is considered on whe
4. electrical variables where n varies from 1 to N being N the number of circuit variables Note that in this case Y n is unique value adopted for all the dX n variable variations 2 local Y n is defined as the maximum absolute value of X n in the time interval till the current time point Note that in this case Y n is different for each electrical variable 3 point local Y n X n I The convergence check tightens as ntoltype is increased Default value of this parameter is 2 local keepmatrixorder Boolean 1 when true the algorithm that LU factorise the Jacobian matrix at each iteration of the Newton method tries to keep the same matrix order In general when the Jacobian matrix is LU factorised the algorithm tries to limit the number of generated fill ins extra extried in the sparse Jacobian matrix and the grouth of the round off error This is accomplished by reordering the rows and the column of the Jacobian matrix Reordering requires the identification of the pivoting entries and submatrix reduction by meas of row column linear combinations This is an extremly cpu time consuming task that often takes a very large portion of the total simulation time mainly in large circuits Experience shows that if the Jacobian matrix is LU factorised by keeping a fixed reodering schema the total simulation time reduces of about 6 times The main drawback is that the grouth of the round off erro
5. format will be used it is very simple and almost free format PAN netlist and control files are usually written using extension pan and simulation execution is in the most simple case started with prompt gt pan netlist pan The few basic rules that any user needs to know in order to write simple netlist files are Comments Comments start from character and continue till the end of line Devices Each device instance is described by a single line starting with the instance name that must obviously be unique then the list of nodes to which the instance is connected the device type e g resistor capacitor etc and finally instance specific parameters and options More details are given in the following Analysis Analysis are in general written on a single line The first token on the line must be a unique identifier the second is the analysis type e g dc tran etc finally all analysis parameters and options Long lines For formatting or aesthetic purposes it is possible to break long lines using character ane Nodes Nodes are implicitly defined by the instance definition lines the only exception is the reference node that must be explicitly defined using the netlist line or card as in the old spice tradition ground electrical name where name can be any unique identifier Options and Parameters device and analysis parameters and options are in general written in the form
6. lcdscd Length dependence of cdscd lcit Length dependence of cit 1lnfactor Length dependence of nfactor 1xj Length dependence of xj lvsat Length dependence of vsat lat Length dependence of at 1a0 Length dependence of a0 lags Length dependence of ags lai Length dependence of al la2 Length dependence of a2 lketa Length dependence of keta 1Insub Length dependence of nsub Indep Length dependence of ndep 1nsd Length dependence of nsd lphin Length dependence of phin 1Ingate Length dependence of ngate lgammai Length dependence of gammal lgamma2 Length dependence of gamma2 lvbx Length dependence of vbx lvbm Length dependence of vbm 1xt Length dependence of xt 1k1 Length dependence of k1 1kt1 Length dependence of kt1 1kt11 Length dependence of kt11 1kt2 Length dependence of kt2 1k2 Length dependence of k2 1k3 Length dependence of k3 1k3b Length dependence of k3b 1w0 Length dependence of w0 ldvtpO Length dependence of dvtp0 ldvtpi Length dependence of dvtp1 1llpeO Length dependence of Ipe0 llpeb Length dependence of Ipeb ldvtO Length dependence of dvt0 ldvt1 Length dependence of dvt1 ldvt2 Length dependence of dvt2 ldvtOw Length dependence of dvtOw ldvtiw Length dependence of dvtlw ldvt2w Length dependence of dvt2w ldrout Length dependence of drout ldsub Length dependence of dsub lvthO Length dependence of vto lvtho Length dependence of vto lua Length dependence of ua luat Length dependence of ual lub Length dependence of ub
7. lvar LVAR lap LAP wvar WVAR wot WOT tr TR vtor VITOR stvto STVTO slvto SLVTO sl2vto SL2VTO sl3vto SL3VTO swvto SWVTO kor KOR slko SLKO 12ko0 SL2KO swko SWKO kr KR slk SLK 62 Chapter 4 Reference s12k SL2K swk SWK phibr PHIBR vsbxr VSBXR slvsbx SLVSBX swvsbx SWVSBX betsq BETSQ etabet ETABET lpi LP1 fbeti FBET1 lp2 LP2 fbet2 FBET2 their THEIR stthetr STTHEIR slthetr SLTHEIR stlthel STLTHE1 gthe1 GTHE1 swthel SWTHE1 wdog WDOG fthe1 FTHE1 the2r THE2R stthe2r STTHE2R slthe2r SLTHE2R stlthe2 STLTHE2 swthe2 SWTHE2 the3r THE3R stthe3r STTHE3R slthe3r SLTHE3R stlthe3 STLTHE3 swthe3 SWTHE3 gamir GAMIR slgami SLGAM1 swgaml SWGAMI1 etadsr ETADSR alpr ALPR etaalp ETAALP slalp SLALP swalp SWALP vpr VPR gamoor GAMOOR slgamoo SLGAMOO sl2gamoo SL2GAMOO etagamr ETAGAMR mor MOR stmo STMO slmo SLMO etamr ETAMR zetir ZETIR etazet ETAZET 4 4 5 4 4 semiconductor device 63 slzeti SLZET1 vsbtr VSBTR slvsbt SLVSBT air AIR stal STAI slai SLAI swal SWA1 a2r A2R sla2 SLA2 swa2 SWA2 a3r A3R sla3 SLA3 swa3 SWA3 tox TOX col COL ntr NTR nfmod NFMOD nfr NFR nfar NFAR nfbr NFBR nfcr NFCR 1IL w W dta DTA th3mod TH3MOD MOSFET level 2 device The mosfet transistor model is the two one of Berkeley SPICE3 ELEMENT DEFINITION Name drain gate source bulk ModelName lt parameter value gt ELEMENT PARAMETERS 1 Real 2 0e 6 channel length w Real 2 0e 6 ch
8. most recently Debian Mint and Ubuntu but PAN has been used without any pain with other Linux flavours including RedHat Gentoo Sabayon and several others A Linux live CD image including the latest version of the simulator and of the demo files is also available This allows quick evaluation without actually installing anything in your computer Using PAN is relatively simple and only requires knowledge of basic Linux terminal com mands and the use of any ASCII editor such as vi As it is netlists are entered in ASCII format but as it will be shown in the sequel thanks to the high compatibility of PAN with well known simulator languages almost any standard netlister can be used as long as some adjustments to the command language are made 2 1 2 Online help and quick reference A quick reference is available with prompt gt pan s a list of keywords appears and further help on a specific topic can be obtained with the command prompt gt pan s topic 10 Chapter 2 A simple PAN walk through Finally command prompt gt pan h outputs to terminal a brief list of command line options Most problems a user may encounter can be solved using the on line help Netlist Files PAN automatically understands different netlist formats this means that the circuit netlist can be written using standard schematic capture software PAN specific commands analysis and options must be added manually In the following examples PAN native
9. set back all sweeping parameters at the values they had at the beginning of the sweeping analysis If set to false parameters are left at values assumed at the end of the sweeping analysis checklc Boolean false check and reduce large conductance trees 1lclocal Boolean false use local algorithm for large conductance reduction safetydigits Integer 0 how many more digits of the ALU should be preserved by the large conductance transformation restart Boolean true restart the DC analysis this means that results of the previous analysis are not used as initial guess saman Boolean false turn on the Newton Samanskii method This method tries to not update and factorise the Jacobian matrix at each iteration of the Newton method The Jacobian that was factorised at a step of the Newton algorithm is keep for several following iterations of Newton algorithm The norm of the RHS vector is monitored and if it does not decrease at each iteration of the Newton Samanskii method the Jacobian is re factorised and the Newton Samanskii method restarted thermalnet Boolean false use the thermal network keepmatrixorder Boolean true when true the algorithm that LU factorise the Jacobian matrix at each iteration of the Newton method tries to keep the same matrix order In general when the Jacobian matrix is LU factorised the algorithm tries to limit the number of generated fill ins extra entries in the sparse Jacobian matrix and the growth o
10. 4 7 4 7 1 4 7 Simulator control pseudo analyses 115 noisefmax Real maximum frequency considered in simulating time domain noise effects The time domain noise analysis is automatically turned on by specifying this parameter seed Integer the seed of the random number generators used in the time domain noise analysis If not specified the seed is taken from the system clock Simulator control pseudo analyses Set Pan Options This statement sets or changes various program control options The options may be set in any order and once set retain their value until reset ANALYSIS DEFINITION Name options lt parameter value gt ANALYSIS PARAMETERS gmin Real 1e 12 conducatnce of hidden resistors in general employed in equivalent models of nonlinear devices for example diodes mosfets bjt to improve relinability of models and convergence properties of numerical algorithms gmax Real 100 maximum value that can be assumed by lt gmin gt during gmin stepping continuation method gground Real le 12 conductance of hidden resistors that connect each circuit node to ground These resistors should aid convergence of numerical algorithms If not given its value is assumed equal to that of the lt gmin gt parameter maxgground Real 100 maximum value that can be assumed by lt gground gt during gground stepping continuation method maxoscillations Real 20 maximum allowed number of oscillations durinrg pseudo tran
11. and v t eigenfunctions Floquet multipliers are ordered from the largest towards the smallest the first lt fmnumber gt are considered devcurr Boolean no the currents of devices for example the drain current of mosfets the anode cathode current of diodes are stored in the output file of the Floquet analysis eigfnorm Boolean false normalise the Floquet eigenfunctions Each eigenfunction is rescaled by the corresponding Floquet exponent this results in a periodic function Note that largely dumped or increasing eigenfunction must be multiplied by relatively large or small scaling factors mainly at the end of the period and this can lead to amplification of truncation errors The result is that the Floquet eigenfunction is no longer period since truncatin errors sum up in an umpredictable way to the correct Floquet eigenfunction fft Boolean no it turns on the computation of Fourier series coefficients of the electrical variables The value of the fundamental frequency is derived from the working period of the circuit that is that given by the user in the non autnomous case or that computed by the shooting algorithm in the autonomous case 4 8 Advanced time domain analyses 13 fftharms The number of harmonics considered during the computation of the Fourier series coefficients This number coincides with that of the computed coefficients fftsamples The number of sampling points adopted to compute the coefficients of the Fourier s
12. comb v timescale 1ms 10us module cnt reg X Y reg 1 0 Count initial begin Count 0 Y i end always posedge X begin Ui display Time realtime Count 0 Count 0 Count 1 Count 0 Count 1 end always Count begin if Count 0 amp amp Count 1 endmodule The simulation is executed exactly as any other analog only circuit prompt gt pan pero pan and the usual directory pero raw containing results for each analysis is created Let us first see results from the transient analysis Tran The two analog state variables i e capacitor voltage and inductor current are represented in the netlist file by node voltage x and current 11 i respectively The third state variable is digital and correspond to the 2 bit register Count defined in the Verilog part of the netlist This register is available as a digital net keyword dignet in the analog part of the netlist as cnt Count where cnt is the Verilog module name This digital net is remapped to the analog net z by the d2a converter D2a2 The conversion transforms the two bit value to a single scalar value in the range defined by v1 0 vh VDD i e in our case to the 4 possible values 0 1 3 2 3 1 The transient behaviour of these state variables is shown in the three panels of Fig 3 2 Following the transient analysis a shooting analysis is performed This analysis finds the steady state solution for the circuit Note that
13. default value is 1 and corresponds to standard no threshold Tarjan splitloop Boolean false try to split too large loops of subcircuits generated during circuit partitioning Splitting is performed by cutting nodes and intruducing artificial voltage sources parsolver Boolean flase turn on off the parallel solver by using multiple threads pardevices Boolean flase turn on off the parallel evaluation of device models by using multiple threads hierckt Boolean flase turn on off the solution of the hierarchical version of the circuit In a conventional analysis a circuit hierarchically organised through subcircuits and instances of subcircuit is first flattened i e a flat version of the netlist is generated and then the circuit is solved If this flag is true the circuit is not flattened and is solved on a subcircuit bases maxabsdeltav Real maximum allowed change of node voltages at each iteration of the Newton algorithm maxreldeltav Real 0 999 maximum allowed relative change of node voltages at each iteration of the Newton algorithm The relative value refers to the largest node voltage by the Newton method If the relative voltage variation exceeds the maximum allowed value the Newton step is suitably shrink to meet this limit The Newton direction is preserved maxdeltat Real max change in temperature allowed per iteration This parameter is effective only during an electro thermal analysis maxdevtemp Real 13
14. digital Boolean false tells that this is a digital macromodel The term digital is improperly used in fact it is considered as digital a vcvs whose output makes discrete voltage transitions at specific time instants or at specific values of the driving signals The output values have not to be necessarily two boolean This type of vcvs are evaluated only once at each time instant of the transient analysis when the Newton algorithm has reached convergence If the digital vcvs does not change its output value it is considered that the Newton algorithm has definitely converged in the other case the vcvs output is updated and the time step is rejected Please have a look of the lt trtime gt options for further information trtime This option has meaning only if the lt digital gt option is set true defauls Ins Digital devices make only discrete variations of their output voltages These variations are checked independently from their magnitudes each time there is a variation between two subsequent integration time instants this variation is forced to happen in less than lt trtime gt value In other words output voltage rise and falling times are ensured to be less than 4 3 2 50 Chapter 4 Reference lt trtime gt This parameter helps the simulator to converge since when there is a transition of a lt digital gt CCVS the output voltage in theory must switch instantaneously From the simulator point of view this means
15. keta Body bias coefficient of non uniform depletion width effect nsub Substrate doping concentration ndep Channel doping concentration at the depletion edge nsd S D doping concentration phin Adjusting parameter for surface potential due to non uniform vertical doping ngate Poly gate doping concentration gammai Vth body coefficient gamma2 Vth body coefficient vbx Vth transition body Voltage vbm Maximum body voltage xt Doping depth k1 Bulk effect coefficient 1 kt1 Temperature coefficient of Vth kt11 Temperature coefficient of Vth kt2 Body coefficient of kt1 k2 Bulk effect coefficient 2 k3 Narrow width effect coefficient k3b Body effect coefficient of k3 wO Narrow width effect parameter dvtpO First parameter for Vth shift due to pocket dvtp1 Second parameter for Vth shift due to pocket lpe0 Equivalent length of pocket region at zero bias lpeb Equivalent length of pocket region accounting for body bias 4 4 semiconductor device 79 dvtO Short channel effect coeff 0 dvti Short channel effect coeff 1 dvt2 Short channel effect coeff 2 dvtOw Narrow Width coeff 0 dvtiw Narrow Width effect coeff 1 dvt2w Narrow Width effect coeff 2 drout DIBL coefficient of output resistance dsub DIBL coefficient in the subthreshold region vthO Threshold voltage vtho Threshold voltage ua Linear gate dependence of mobility ual Temperature coefficient of ua ub Quadratic gate dependence of mobility ubi Temperature coefficient
16. lub1 Length dependence of ub1 luc Length dependence of uc luc1 Length dependence of uc1 4 4 semiconductor device 85 lud Length dependence of ud ludi Length dependence of ud1 lup Length dependence of up 1lp Length dependence of Ip 1u0 Length dependence of u0 lute Length dependence of ute lvoff Length dependence of voff 1lminv Length dependence of minv ldelta Length dependence of delta lrdsw Length dependence of rdsw lrsw Length dependence of rsw lrdw Length dependence of rdw lprwg Length dependence of prwg lprwb Length dependence of prwb lprt Length dependence of prt letaO Length dependence of eta0 letab Length dependence of etab lpclm Length dependence of pclm lpdiblci Length dependence of pdiblc1 lpdiblc2 Length dependence of pdiblc2 lpdiblcb Length dependence of pdiblcb lfprout Length dependence of pdiblcb lpdits Length dependence of pdits lpditsd Length dependence of pditsd lpscbe1i Length dependence of pscbel lpscbe2 Length dependence of pscbe2 lpvag Length dependence of pvag lwr Length dependence of wr ldwg Length dependence of dwg ldwb Length dependence of dwb 1b0 Length dependence of bO 1bi Length dependence of b1 lcgsl Length dependence of cgsl lcgdl Length dependence of cgdl lckappas Length dependence of ckappas lckappad Length dependence of ckappad lcf Length dependence of cf 1lclc Length dependence of clc lcle Length dependence of cle lalphaO Length dependence of alphaO lalphai Length de
17. pvthO Cross term dependence of vto pvtho Cross term dependence of vto pua Cross term dependence of ua puai Cross term dependence of ual pub Cross term dependence of ub pub1 Cross term dependence of ub1 puc Cross term dependence of uc puci Cross term dependence of uc1 pud Cross term dependence of ud pudi Cross term dependence of ud1 pup Cross term dependence of up plp Cross term dependence of Ip pu0 Cross term dependence of u0 pute Cross term dependence of ute pvoff Cross term dependence of voff pminv Cross term dependence of minv pdelta Cross term dependence of delta prdsw Cross term dependence of rdsw prsw Cross term dependence of rsw prdw Cross term dependence of rdw pprwg Cross term dependence of prwg pprwb Cross term dependence of prwb pprt Cross term dependence of prt petaO Cross term dependence of eta0 petab Cross term dependence of etab ppclm Cross term dependence of pclm ppdiblc1 Cross term dependence of pdiblc1 ppdiblc2 Cross term dependence of pdiblc2 ppdiblcb Cross term dependence of pdiblcb pfprout Cross term dependence of pdiblcb ppdits Cross term dependence of pdits ppditsd Cross term dependence of pditsd ppscbei Cross term dependence of pscbel ppscbe2 Cross term dependence of pscbe2 ppvag Cross term dependence of pvag pwr Cross term dependence of wr pdwg Cross term dependence of dwg pdwb Cross term dependence of dwb pbO Cross term dependence of b0 pbi Cross term dependence of b1 4 4 semiconductor d
18. vmax Carrier saturation velocity BIN theta Mobility modulation coefficient BIN mss Vdse transition parameter parameter BIN kss Fractions of the channel resistance coefficient rsh Source drain diffusion sheet resistance capmod Intrinsic charge model zeroc Zero gate source gate drain capacitance flag Cgs Cgd 0 if zeroc 1 and capmod 1 4 5 8 102 Chapter 4 Reference intdsnod Intrinsic source and drain nodes usage flag minr Minimum source drain resistance tnom Parameter measurement temperature trise Temperature rise from ambient shmod Self heating selector cthO Self heating thermal capacitance rthO Self heating thermal resistance wthO Minimum width for thermal resistance calculation tmax Maximum device temperature above ambient wow 11 nrd nrd nrs nrs wmax Maximum channel width for which the model is valid wmin Minimum channel width for which the model is valid lmax Maximum channel length for which the model is valid 1min Minimum channel length for which the model is valid acm Area calculation method xl Accounts for masking and etching effects xw Accounts for masking and etching effects 1d Lateral diffusion into channel from source and drain diffusion wd Lateral diffusion into channel from bulk along width 1m1t Length diffusion layer shrink reduction factor wmlt Width diffusion layer shrink reduction factor ldif Length of heavily doped diffusion adjacent to gate BIN hdif
19. 2 Ve Vg 10 12 12 1 1 05 1 1 1 15 1 0792 1 0792 1 0792 time x 10 time 10 Figure 2 15 Details of the long envelope analysis for the Butler oscillator towards the end of the simulation Ve Vg 10 Sh emitter Sh base 12 l i 1 1458 1 1458 1 1458 time x1 0 Figure 2 16 Final shooting analysis for the Butler oscillator 2 3 AMS simulation with PAN Modern circuit design is often based on the Analog Mixed Signal AMS paradigma where various parts of the overall system are modeled using a different approach The most common situation is where an analog circuit such as the ones seen in the previous section are connected with digital circuits or to blocks described using behavioural languages PAN is capable of simulating AMS circuits seamlessly integrating the analog netlist with a Verilog description of the digital part and a Verilog A description of the behavioural part Simple analysis are available but time domain steady state or more complex analysis such as PAC pr PNOISE are also available This is not common since standard steady state algorithms require continuous vector fields and this is not the case of AMS circuits In the following a few simple examples mostly derived from those published by the authors in Biz are shown from an operative point of
20. 2 A simple PAN walk through since Trani starts from time t 100s total simulation time is 20s Solutions i e last points of time domain analysis can also be saved and loaded using commands load lt filename gt and save lt filename gt respectively Results are shown in Fig 2 8 as it can be seen since the previous analysis ends up very close to the oscillator steady state trajectory the circuit limit cycle is evidenced A simple linear LC oscillator and frequency warping effects The circuit we consider now has the sole purpose of showing a little known effect on simulation of the integration method employed BS03 PAN default integration method is trapezoidal but also the Gear methods up to order 6 are implemented Consider the netlist in 2 3 2 Netlist 2 3 2 Linear LC oscillator File 1c pan ground electrical 0 Analyses TranO tran tstop 1000 uic 1 default trapezoidal method Trani tran tstop 1000 uic 1 method 2 order 2 Gear order 2 Tran2 tran tstop 1000 uic 1 method 2 order 6 Gear order 6 3 Circuit cl n10 O capacitor c 1 icon 3 11 n10 O inductor 1 1 Three time domain analyses are performed using three different integration methods Results are quite interesting All simulations start from the same initial conditions see Fig 2 9 left N wo Ff gt 0 gt 1 2 3 H 4 1 1 i i 0 5 10 15 20 700 705 710 715 720 time time Figure 2 9 Capa
21. Integer 3 degree of annotation Values from 1 to 7 turn on displaying of generic informations about the analysis run Values above 7 are bitwise and turn on specific printouts and in particular 0x08 print some information about the convergence of the Newton method 0x10 print MNA matrix and RHS vector at each iteration of the Newton method 0x20 at each iteration of the Newton method the MNA matrix is written in a file named as the analysis 0x40 write the Floquet matrix in a file named as the analysis time Output time domain signals maxiters Integer 100 allowed maximum number of iterations restart Boolean true restart the analysis that is do not use previous solution as initial guess printnodes Print internal nodes numbers solver Integer 0 the solver employed to compute the solution Available solvers are 0 direct solver block LU factorisation 1 iterative solver gmres ta Boolean false this parameter turns on and off the time assisted harmonic balance solver method Select the integration method used in the time assisted harmonic balance 1 trapezoidal default 2 Gear method of various order order Set the order of the Gear integration method used in the time assisted harmonic balance default value is 2 It has no meaning if other integration methods are employed load String the path of the file from which a previously saved solution is read The file format can be both ascii and bz2 and it is automatica
22. Length dependence of moin lnoff Length dependence of noff lvoffcv Length dependence of voffcv wcdsc Width dependence of cdsc wcdscb Width dependence of cdscb wcdscd Width dependence of cdscd wcit Width dependence of cit wnfactor Width dependence of nfactor wxj Width dependence of xj wvsat Width dependence of vsat wat Width dependence of at wa0 Width dependence of a0 wags Width dependence of ags wail Width dependence of al wa2 Width dependence of a2 wketa Width dependence of keta wnsub Width dependence of nsub wnch Width dependence of nch wngate Width dependence of ngate wgammal Width dependence of gammal wgamma2 Width dependence of gamma2 wvbx Width dependence of vbx wvbm Width dependence of vbm wxt Width dependence of xt wk1 Width dependence of k1 wkt1 Width dependence of kt1 wkt11 Width dependence of kt11 wkt2 Width dependence of kt2 4 4 semiconductor device 73 wk2 Width dependence of k2 wk3 Width dependence of k3 wk3b Width dependence of k3b wwO Width dependence of w0 wnlx Width dependence of nlx wdvt0O Width dependence of dvtO wdvt1 Width dependence of dvt1 wdvt2 Width dependence of dvt2 wdvtOw Width dependence of dvt0w wdvtiw Width dependence of dvtlw wdvt2w Width dependence of dvt2w wdrout Width dependence of drout wdsub Width dependence of dsub wvtho Width dependence of vto wvthO Width dependence of vto wua Width dependence of ua wua1 Width dependence of ual wub Width dependence of ub wub1
23. Length of heavily doped diffusion from contact to lightly doped region xj Metallurgical junction depth meto Fringing field factor for gate to source and gate to drain overlap capacitance rdc Additional drain resistance due to contact resistance rsc Additional source resistance due to contact resistance dvt The difference between von and the threshold BIN vsi Above threshold DIBL parameter BIN vst Above threshold DIBL parameter BIN binunit Bin parameter unit selector 1 for microns and 2 for meters binflag Binflag 2 is to open xl xw binning Time controlled switch The switch is tied between n1 and n2 nodes and it is driven by the time sequense specified by means of the ton and toff time instants There is no upper limit the the number of the ton and toff values The resistance of the switch is set to lt ron gt when it is closed and to lt roff gt when open ELEMENT DEFINITION Name nl n2 ModelName lt parameter value gt ELEMENT PARAMETERS region Integer 0 working status of the switch at the beginning of the time sequence 1 on 0 off For example if the status is on the time sequence must start with a lt toff gt value that turns off the switch followed by a lt ton gt value that turns on the switch and so on ton Real time instant at which the switch is tuned on There can be several of these time 4 5 9 4 5 misc device 103 instants toff Real time instant at which the switch is tuned
24. PARAMETERS capmod Capacitance model selector diomod Diode IV model selector rdsmod Bias dependent S D resistance model selector trngsmod Transient NQS model selector acngsmod AC NQS model selector mobmod Mobility model selector rbodymod Distributed body R model selector rgatemod Gate R model selector permod Pd and Ps model selector geomod Geometry dependent parasitics model selector 78 Chapter 4 Reference fnoimod Flicker noise model selector tnoimod Thermal noise model selector igcmod Gate to channel Ig model selector igbmod Gate to body Ig model selector tempmod Temperature model selector paramchk Model parameter checking selector binunit Bin unit selector version parameter for model version toxe Electrical gate oxide thickness in meters toxp Physical gate oxide thickness in meters toxm Gate oxide thickness at which parameters are extracted toxref Target tox value dtox Defined as toxe toxp epsrox Dielectric constant of the gate oxide relative to vacuum cdsc Drain Source and channel coupling capacitance cdscb Body bias dependence of cdsc cdscd Drain bias dependence of cdsc cit Interface state capacitance nfactor Subthreshold swing Coefficient xj Junction depth in meters vsat Saturation velocity at thom at Temperature coefficient of vsat a0 Non uniform depletion width effect coefficient ags Gate bias coefficient of Abulk a1 Non saturation effect coefficient a2 Non saturation effect coefficient
25. Real 0 0 the magnutude of the voltage probe vprobephi Real 0 0 the phase of the voltage probe vprober1 Real 0 0 the value of the resistor connected in series to the voltage probe vprobemax Real the maximum value of the magnitude of the voltage probe employed during the sweep to find the steady state solution iprobemax Real the maximum value of the magnitude of the current probe employed during the sweep to find the steady state solution gminstepping Boolean flase turn on and off the gmin stepping continuation method ggroundstepping Boolean flase turn on and off the gground stepping continuation method srcstepping Boolean flase turn on and off source stepping initstep Init source stepping value minstep Minimum step allowed in source stepping ireltol Relative current convergence criterion iabstol Absolute current convergence criterion vreltol Real 1 0e 6 relative voltage convergence criterion vabstol Real 1 0e 6 absolute voltage convergence criterion Almost Periodic Steady State Analysis The analysis computes the periodic steady state solution of a circuit Both non autonomous and autonomous circuits are supported Results made available to the lt control gt analysis are voltage node and branch current spectra output spectra employed fundamental and beats keyword to access them is lt freq gt matrix of indices for the retrival of fundamentals and beats inside the lt freq gt vec
26. Vdsat parameter for C V model cle Vdsat parameter for C V model 82 Chapter 4 Reference dwc Delta W for C V model dlc Delta L for C V model xw W offset for channel width due to mask etch effect xl L offset for channel length due to mask etch effect dlcig Delta L for Ig model dwj Delta W for S D junctions alphaO substrate current model parameter alphai substrate current model parameter beta0O substrate current model parameter agidl Pre exponential constant for GIDL bgidl Exponential constant for GIDL cgidl Parameter for body bias dependence of GIDL egidl Fitting parameter for Bandbending aigc Parameter for Igc bigc Parameter for Igc cigc Parameter for Igc aigsd Parameter for Igs bigsd Parameter for Igs cigsd Parameter for Igs aigbacc Parameter for Igb bigbacc Parameter for Igb cigbacc Parameter for Igb aigbinv Parameter for Igb bigbinv Parameter for Igb cigbinv Parameter for Igb nigc Parameter for Igc slope nigbinv Parameter for Igbinv slope nigbacc Parameter for Igbacc slope ntox Exponent for Tox ratio eigbinv Parameter for the Si bandgap for Igbinv piged Parameter for Igc partition poxedge Factor for the gate edge Tox ijthdfwd Forward drain diode forward limiting current ijthsfwd Forward source diode forward limiting current ijthdrev Reverse drain diode forward limiting current ijthsrev Reverse source diode forward limiting current xjbvd Fitting parameter for drain diode breakdown current xjbvs F
27. a is a scalar and b a vector a is compared with each entry of b and the minimum scalar value is returned in x The same happens if b is a matrix The role of a and b can be exchanged If a and b are both vectors x is a vector containing the minimum between the the corresponding entries of a and b If a and b are matrices x is a matrix containing the minimum between the corresponding entries of a and b For complex number the norm is considered x mktime tm convert a time structure corresponding to the local time to the number of seconds since the epoch x err iters multirootf func x deriv iters abstol find a solution of a system of nonliear equations by adopting a hybrid version of the Newton iterative method The r func x function takes as argument the x vector representing the initial value of the solution and returns the corresponding value of the r residue The j deriv x function takes as input the current value of the solution and returns the Jacobian matrix If deriv is not specified argument is omitted since multirootf is called with only two arguments or the third argument is equal to null the Jacobian matrix is automatically estimated by the Newton hybrid method The iter variable is optional and defines the maximum allowed number of iterations The abstol variable is opti
28. a point in the pwl sequence It is employed only in time domain analyses If the simulation time stop is set after the last time point of the pwl sequence 4 2 5 4 2 Basic one port device 47 the entire sequence is repeated as a periodic signal tr Real 0 0 rising time of the pulse generator trise Real 0 0 rising time of the pulse generator rise Real 0 0 rising time of the pulse generator tf Real 0 0 falling time of the pulse generator fall Real 0 0 falling time of the pulse generator tfall Real 0 0 falling time of the pulse generator vi Real 0 0 voltage value of the pulse waveform in the time interval outside that specified by the duty cycle val0O Real 0 0 voltage value of the pulse waveform in the time interval outside that specified by the duty cycle v2 Real 0 0 voltage value of the pulse waveform in the time interval specified by the duty cycle vali Real 0 0 voltage value of the pulse waveform in the time interval specified by the duty cycle td Real 0 0 time delay before the first rising front in the waveform generated by the pulse generator width Real 0 0 duty cycle of the pulse waveform It must be an absolute time interval and not a relative value with respect to the period of the pulse generator period Real 0 0 the period of the pulse and pwl generator If not specified for a pwl generator then the waveform is continued above the last time point by maintaining the last volt
29. are retrieved If index is a negative number the last index elements of the result vector are retrieved x gmres func y b k the gmres method solves the A x b linear equation system The func function performs the A x matrix by vector product The y vector is the initial guess The k parameter is optional and specifies after how many iterations the gmres method is restarted x ifft y computes the inverse fast Fourier transform of y The y argument can be a vector or a matrix in this case the Fourier transform is performed column by column and the x result is a matrix imag z integral y x computes the integral of vector y versus vector x The lengths of these vectors must be the same The z returned vector has the same dimension of x and stores the value assumed by the integral versus x y interp1 x0 y0 x method interpolates values where x0 stores the known x values y0 stores the known y values x stores the desired value s The length of y0 must be equal to that of x0 The method argument is optional allowed values are linear or bspline inv inverse of a matrix y isnan x returns logical variable array matrix which is true where the elements of x are NaN values and false where they are not For example isnan 13 Inf NaN gt 0 0 1 1 F length l u p
30. can be computed with t1 t2 sqrt L1 L2 K can be calculated from the L1 the inductance of winding 1 with winding 2 open and Ls the inductance of winding 1 with winding 2 shorted with K sqrt 1 Ls L1 t1 t2 o o o Lel o wl o Ill o w2 o Le2 Ill Lm Ill 1 Ill o wl o Ill o w2 4 3 6 4 3 7 4 3 8 4 3 Basic two port device 55 ELEMENT DEFINITION Name pl pl s2 s2 transformer lt parameter value gt ELEMENT PARAMETERS t1 Number of turns on winding 1 t2 Number of turns on winding 2 Ideal Nullor Nullor is a two port device port 1 is identified by the first two nodes of the device and port 2 by the remaining two nodes Its characteristic at port 1 is described by the il branch current and v1 branch voltage Both are always null il 0 and v1 0 The characteristic of port 1 is composed of only a point located in the origin of the voltage current axes nullator Port 2 is described by current i2 and voltage v2 they are not related and can independently assume any value norator ELEMENT DEFINITION Name pl pl s2 s2 nullor Coupled Inductors This instance is not a true one but it allows coupling of two or more inductors It must be specified the name of at least two inductor instances through a parameter and the corresponding coupling coefficient If an inductor is coupled to more than one other inductor a sequence of inductance instances must be specified with the co
31. chosen to keep variation with respect to total stored energy below the user defined threshold In general this method gives less accurate results with respect to the previous one but the analysis runs faster etoltype Integer 2 this option specifies the tolerance check adopted to check if the envelope analysis has reached convergence inside a simulation window one working period of the circuit Checks are done on the variation of variables computed by the time domain analysis along one time window Define as dX n the variation of the n electrical variable between the beginning and the end of one time window as reltol and abstol the relative and absolute maximum allowed variations respectively The n th variable has converged if dX n lt reltol Y n abstol This parameter defines three different ways to compute Y n numbered as 1 2 and 3 Their meanings are listed in the following 1 global Y n is defined as the maximum over one all time windows till the current one of the maximum value assumed by the X n electrical variables where n varies from 1 to N being N the number of circuit variables Note that in this case Y n is unique value adopted for all the dX n variable variations 2 local Y n is defined as the maximum absolute value of X n in the time window Note that in this case Y n is different for each electrical variable 3 point local
32. damped version of the Newton method In general this increses robustnes of analyses by avoiding drawbacks such as overflows if for example exponential functions are used in the macro model file String S parameter data file name scale Real 1 0 frequency scale factor veriloga String list the path s of the file s that must be compiled in order to implement an lt veriloga gt module adms String list the path s of the file s that must be compiled in order to implement an lt veriloga gt module verilogapp Boolean false the output of the lt veriloga gt pre processor is saved in a file named as the module followed by the pp suffix verilogamodname String the name of the veriloga module linked to the lt nport gt model If this parameter is omitted the module with the same name of the lt nport gt model is looked for verilogaelectrical Boolean true automatically add the lt electrical gt discipline if it is not specified in the veriloga file verilogashadowvar Boolean true a warning is given when a local variable shadows a previous one with same name macro Boolean false the device is described by an external macromodel loaded as a shared 4 5 6 4 5 misc device 99 executable module module String the name file path of the shared executable implementing the macromodel evaluate String the name of the routine in the shared executable that evaluates the macro model args Stri
33. in periodic small signal analysis pss analysis pacphase Real 0 0 phase of the source in small signal periodic analysis pac analysis magit Real 0 0 magnitude of first fundamental in spectral analysis mag2t Real 0 0 magnitude of second tone generated in the spectral analysis freq2t List the list of indices that defines the frequency of the second tone as a linear combination of the two fundamentals of the spectral analyses It is composed of two items The fist one is the index of the harmonic of the first fundamantal and the second is the index of the harmonic of the second fundamental For example freq2t 1 1 defines the frequency 1 f1 1 f2 phaseit Real 0 0 phase of second time in the spectral analysis phase2t Real 0 0 phase of second time in the spectral analysis i Current corner of the pwl time function t Time instant of a corner in the pwl function Pwl sequence is considered only in time domain analyses tr Rising time of the pulse waveform tf Falling time of the pulse waveform i2 Current value during the up portion of the pulse waveform Note that this value may be negative i1 Current value during the low portion of the pulse waveform td Real 0 0 time delay before the first rising front in the waveform generated by the pulse generator or the time delay before the beginning of the first period of the sinusoidal waveform generated by the sinusoidal generator width Duty cycle of the pulse
34. insertion technique is used printnodes Print internal nodes numbers restart Boolean true or String restart the analysis that is do not use previous solution as initial guess In the other case boolean false the result from the previous analysis are used as initial guess If the name of a previous analysis is given the results computed by this analysis are used as initial guess therefore it can be not necessarily the previous one load String the path of the file from which a previously saved solution is read The file format can be both ascii and bz2 and it is automatically determined during loading This solution is used as initial guess to the spectral algorithm save String the path of the file into which the solution is saved bz2 format is adopted The solution stored in this file can be later loaded by the simulator and used as initial condition from which the spectral algorithm is started saman Real 2 0 samanskii constant At each iteration of the Newton algorithm the norm of the residue right hand side vector is computed and compared to that computed at the previous iteration If the ratio between the two norms is less than the value of the samanskii constant plus one then the Jacobian is recomputed In general Jacobian recomputation speeds up convergence srcstepping Apply source stepping minstep Minimum allowed step in source stepping maxiter Max number of iterations gminstepping Boolean flase turn on a
35. is shown in Fig 2 2 2 i 1 1 i i 100 80 60 40 20 0 20 Vy Figure 2 2 Diode current vs generator voltage generated by Netlist 2 2 4 2 2 3 The alter pseudo analysis Even if alter is listed among the analysis and follows the same syntax it is actually a directive instructing the simulator to modify a parameter or option value The typical use of this statement is that of performing the same analysis with different values of circuit or device parameters or 18 Chapter 2 A simple PAN walk through global options Consider once more the diode DC sweep example but assume we are interested in comparing its characteristic at different temperatures A simple way to do this is shown in Netlist 2 2 5 Netlist 2 2 5 Diode DC sweep with alter statements File Simple_003 pan ground electrical 0 Analyses DcO dc print yes instance vil param vdc start 85 0 step 0 01 stop 10 0 Alter alter opt yes param temp value 10 Dc1 dc print yes instance vi param vdc start 85 0 step 0 01 stop 10 0 Alter2 alter opt yes param temp value 120 Dc2 dc print yes instance vi param vdc start 85 0 step 0 01 stop 10 0 3 Circuit vi ni 0 vsource vdc 2 di ni 0 mydiode Models model mydiode diode is 3 142f rs 3 536 n 1 xti 3 eg 1 11 m 0 4186 vj 0 75 fc 0 5 bv 80 ibv 24 573m imax 100 Three subsequent DC analysis are performed at three different temperatures the first at nominal temp
36. is swept or the frequency value in single frequency analysis This parameter must not be specified contemporary to the frequency sweeping command print Boolean false print the result in single frequency analysis start Start sweep limit stop Stop sweep limit center Center of sweep span Sweep limit span step Step size linear sweep lin Number of points linear sweep dec Points per decade log Number of points log sweep sweep Device whose parameter value is to be swept param Name of parameter to be swept Time domain analysis The analysis computes the time domain behavior of linear and nonlinear circuits The initial conditions are computed automatically through a DC analysis or can be specified by the user In this case the user gives the initial voltage across capacitors and current through inductors Time domain integration employs different types of linear multi step methods such as backward Euler trapezoidal and Gear methods There are two different mechanisms to control the integration time step Waveforms made available to control analysis are 1 Node voltages and branch currents computed by transient analysis time samples are stores in the time vector Waveforms can be accessed through the command get TranName node 2 Spectra of time domain node voltage and branch current waveforms frequency samples are stored in the freq vector The spectrum of a node voltage and branch current is named as the el
37. ldsub Length dependence of dsub lvtho Length dependence of vto lvthO Length dependence of vto lua Length dependence of ua luat Length dependence of ual lub Length dependence of ub lub1 Length dependence of ub1 luc Length dependence of uc luc1 Length dependence of ucl 1u0 Length dependence of u0 lute Length dependence of ute lvoff Length dependence of voff lelm Length dependence of elm ldelta Length dependence of delta lrdsw Length dependence of rdsw lprwg Length dependence of prwg lprwb Length dependence of prwb lprt Length dependence of prt letaO Length dependence of eta0 letab Length dependence of etab lpclm Length dependence of pclm lpdiblc1 Length dependence of pdiblc1 lpdiblc2 Length dependence of pdiblc2 72 Chapter 4 Reference lpdiblcb Length dependence of pdiblcb lpscbe1 Length dependence of pscbe1 lpscbe2 Length dependence of pscbe2 lpvag Length dependence of pvag lwr Length dependence of wr ldwg Length dependence of dwg ldwb Length dependence of dwb 1b0 Length dependence of bO 1bi Length dependence of b1 lcgsl Length dependence of cgsl lcgdl Length dependence of cgdl lckappa Length dependence of ckappa lcf Length dependence of cf 1lclc Length dependence of clc 1lcle Length dependence of cle lalphaO Length dependence of alphaO lalphai Length dependence of alphal lbeta0 Length dependence of beta0 lvfbcv Length dependence of vfbcv lvfb Length dependence of vfb lacde Length dependence of acde 1lmoin
38. list In this case if the spectrum of the solution computed by the previous shooting analysis has been determined the noise computed by the pnoise analysis is displayed both in dB and in dBc The pnoise analysis can also decompose the computed total noise in the phase and amplitude components The phase noise is defined as the projection of the total noise along the eigenfunction corresponding to the Floquet multiplier theoretically equal to 1 The amplitude 136 Chapter 4 Reference noise is defined as the difference between the total noise and the phase noise Waveforms that can be stored in memory and made available to the control analysis are the computed noise spectrum and its phase and amplitude components These vectors can be retrieved through the statements get AnalysisName out get AnalysisName out pm and get AnalysisName out am The frequency axis can be retrived through the stateemnt get AnalysisName freq ANALYSIS DEFINITION Name pnoise lt parameter value gt ANALYSIS PARAMETERS annotate Degree of annotation It is an integer number values from 1 to 7 turn on displaying of generic informations about the analysis run Values above 7 are bitwise and turn on specific printouts 0x08 printout the state transition matrix 0x10 save the state transition matrix in a file named as the analysis start Start sweep limit stop Stop sweep limit center Center of sweep span
39. lu a decomposes a square matrix in the l lower triangular matrix the u upper triangular one and the p permutation one such that we have 1 u pa x max a if 120 Chapter 4 Reference a is a vector x is the scalar maximum entry of a If a is a matrix x is a vector whose elements are the maximum entries of the corresponding column of a For example the second entry of x stores the maximum entry in the second column of a x max a b a and b can be scalars vectors matrices If a is a scalar and b a vector a is compared with each entry of b and the maximum scalar value is returned in x The same happens if b is a matrix The role of a and b can be exchanged If a and b are both vectors x is a vector containing the maximum between the corresponding entries of a and b If a and b are matrices x is a matrix containing the maximum between the corresponding entries of a and b mean x min a if a is a vector x is the scalar minimum entry of a If a is a matrix x is a vector whose elements are the minimum entries of the corresponding column of a For example the second entry of x stores the minimum entry in the second column of a x min a b a and b can be scalars vectors matrices If
40. mee in Eea yli I Se ve A ez ii e fi salty eo azli E EAT A AIN RDN MATE ea fili pn yee en a ave EN A say nS g a E ai E ARIE nA IE Reedy sai A Fa Ne ee Sao Le Tas ase M pe ge ye rS SR ee aa Soe Ro 1 eat ir AYI tS pa bot aie ii anes vi VAO ME EEN ye a EN RO TRUE IA ott the ae eee Resse f 3 TOE lie mst en tks a er G ETES O VE UUL EN ESU oh cee TSS Se OS Vm R wa ae ts 2 1 Overview The PAN simulator is mainly a circuit simulator but can be easily extended and controlled in order to perform several different types of system simulations PAN is capable of analog digital or AMS simulation electrothermal simulation and can be extended by linking external modules A complete list of anayses devices and directives that are available natively i e without user extensions are listed in the reference section at the end This chapter represents a sort of tutorial showing in the most direct way and through a limited number of examples how to use the PAN circuit simulator to perform simple simulation tasks More complex tasks and a detailed definition of all parameters proper to each analysis method will be given in later chapters 2 1 1 Where and How downloading and installing The PAN circuit simulator binary can be downloaded from http brambilla ws dei polimi it The PAN binary requires a standard linux installation PAN developers use Debian based systems
41. name value separated by spaces Boolean values can be entered also using keywords yes or no that correspond respectively to 1 and 0 More complex rules such as those needed to define subcircuits models or control structures will be shown later These simple rules are exemplified below in Netlist 2 1 1 Netlist 2 1 1 Basic netlist writing rules ground electrical 0 This is a comment 3 Circuit ri ni n2 resistor r 1k a resistor instance InstanceName Node1 Node2 Node3 Node4 2 1 Overview 11 Node5 Node6 Node7 MultiportDevice option1 10 option2 30 Instance of MultiportDevice with line breaks Models and devices Any circuit is essentially described by a topology and a set of components Components can be defined in many different ways PAN includes a large number of predefined standard devices several customizable devices and the possibility to define completely new devices by writing code in many different languages Moreover any device predefined or user defined can be made in to a template using the mode1 statement This is a standard feature of most spice like simulators Netlist 2 1 2 An example of definition and use of a diode model ground electrical 0 3 Circuit ri ni 0O resistor r 1k di ni 0O MYDIODE d2 ni 0 MYDIODE bv 100 imax 100 Models model MYDIODE diode is 3 142f rs 4 2 bv 80 An example of model definition and use is shown in Netlist 2 1 2 n
42. not given all state variables are monitored solver Integer 0 the employed solver available ones are 0 direct solver based on the Newton method and the LU factorisation of the fundamental matrix 1 direct solver based on the Newton method and the QR factorisation of the fundamental matrix 2 direct solver based on hybrid Newton method 3 iterative solver based on the Newton method and the gmres method As simple guidelines to choose a solver one has to consider that in general the direct solvers are faster on small and medium size circuits the iterative solver largely outperformes the direct one on large circuits Circuits with more than a thousand nodes are considered as large in this case the iterative solver is automatically chosen if this option is not specified that is if a solver is not directly specified by the user through this option The iterative solver requires tens time less memory than direct solvers If direct solvers are chosen they take a smaller amount of memory if the Gear integration method is employed The QR factorisation method of the fundamental matrix is preferred to the LU one since there are stiff circuits where some pseudo state variables yield Floquet multipliers extremely close to 1 These Floquet multipliers indicate that the fundamental matrix is singular Singularities can be handled by QR factorisation since the columns of the Q matrix can span the subspace into which the solution lives If
43. of the monodromy matrix eigf Boolean false the periodic eigenfunctions of the Floquet analysis are computed In particular the uj t and vj t eigenfunctions related to the pseudo state variables of the MNA formulation are computed Furthermore the vjB t functions that relate possible small signal perturbations applied to the circuit nodes to the pseudo state variables and the Muj t eigenfunctions that relate the pseudo state variables to the node potentials and device branch currents of those elements that are not voltage controlled are also computed samples Integer 200 the number of samples used to compute the sensitivity functions Floquet eigenfunctions with respect to pseudo state variables along one working period of the circuit Floquet eigenfuncitons are computed in a reduced number of time points with respect to those used by the variable time step length integration algorithms employed by the time domain analysis This is done mainly to reduce the total ammount of memory needed to store and elaborate the results The minimum allowed number of samples is equal to 100 printmo Boolean no when true the fundamental matrix is printer In case of autonomous circuits the fundamental matrix is bordered with an extra column and an extra row that refer to the of the solution with respect to the working period of the circuit fmnumber Integer 32766 The maximum number of the Floquet multipliers that are considered for computing the u t
44. of ub uc Body bias dependence of mobility uci Temperature coefficient of uc ud Coulomb scattering factor of mobility udi Temperature coefficient of ud up Channel length linear factor of mobility lp Channel length exponential factor of mobility u0 Low field mobility at Tnom eu Mobility exponent ute Temperature coefficient of mobility voff Threshold voltage offset minv Fitting parameter for moderate invversion in Vgsteff voff1 Length dependence parameter for Vth offset tnom Parameter measurement temperature cgso Gate source overlap capacitance per width cgdo Gate drain overlap capacitance per width cgbo Gate bulk overlap capacitance per length xpart Channel charge partitioning delta Effective Vds parameter rsh Source drain sheet resistance rdsw Source drain resistance per width rdswmin Source drain resistance per width at high Vg rsw Source resistance per width rdw Drain resistance per width rdwmin Drain resistance per width at high Vg rswmin Source resistance per width at high Vg prwg Gate bias effect on parasitic resistance prwb Body effect on parasitic resistance prt Temperature coefficient of parasitic resistance eta0 Subthreshold region DIBL coefficient etab Subthreshold region DIBL coefficient pclm Channel length modulation Coefficient pdiblci Drain induced barrier lowering coefficient pdiblc2 Drain induced barrier lowering coefficient pdiblcb Body effect on drain induced barrier lowering fprout Rout degradation co
45. on page 12 Articles BBS11 F Bizzarri A Brambilla and G Storti Gajani Phase Noise Simulation in Analog Mixed Signal Circuits An Application to Pulse Energy Oscillators In Circuits and Systems II Express Briefs IEEE Transactions on 58 3 2011 pages 154 158 ISSN 1549 7747 DOI 10 1109 TCSII 2011 2111570 cited on page 35 Biz Federico Bizzarri et al Simulation of real world circuits extending conventional analysis methods to circuits described by heterogeneous languages In Circuits and Systems Magazine IEEE Forthcoming cited on page 35 BGS09 A Brambilla G Gruosso and G Storti Gajani Determination of Floquet Exponents for Small Signal Analysis of Nonlinear Periodic Circuits In Computer Aided Design of Integrated Circuits and Systems IEEE Transactions on 28 3 2009 pages 447 451 ISSN 0278 0070 DOI 10 1109 TCAD 2009 2013285 cited on page 32 146 Chapter 4 Reference BGS 10 BMS05 BS03 Bra 10 HRB75 A Brambilla G Gruosso and G Storti Gajani FSSA Fast Steady State Algorithm for the Analysis of Mixed Analog Digital Circuits In Computer Aided Design of Integrated Circuits and Systems IEEE Transactions on 29 4 Apr 2010 pages 528 537 DOI 10 1109 TCAD 2010 2042886 cited on page 32 A Brambilla P Maffezzoni and G Storti Gajani Computation of period sensitivity functions for the simulation of phase noise in oscillat
46. paramenter in particular 1 it is assumed that all the capacitors and inductors have suitable initial conditions This means that if they are not specified by the user they are assumed equal to zero Note that this initial condition determination does not deals with loops of voltages source and capacitors even though consistent initial conditions are given since the loop current is undertermined The same applies to inductor cut sets 2 it is assumed that all the capacitors and inductors whose initial conditions have not been specified are open circuit and short circuit respectively In this case suitable voltage source and capacitor loops orcurrent source and inductor cut sets do not cause any problem Steady state computation through shooting method Shooting analysis it is a time domain analysis that efficiently determines the steady state behaviour of both autonomous and nonautonomous circuits When dealing with autonomous circuits the working period is automatically computed by starting from an initial guess given by the user at the beginning of the analysis When a limit cycle of an autonomous circuit has been found it is possible to compute the Floquet multipliers that give insight to the local stability of the limit cycle When the annotation level is suitably set at each iteration of the algorithm some usefull information is printed For example consider the following line Il R ll 12 992 D 10 R 10 F 5 00000000 MHz lt 4 gt M
47. periodic small signal analysis pac analysis only the xacmag and xacphase parameters are used DC voltage and magnitudes are given in volts and phase is given in degrees ELEMENT DEFINITION Name n n vsource lt parameter value gt ELEMENT PARAMETERS vdc Real 0 0 dc voltage dc Real 0 0 dc voltage stepping Boolean true enables or disable the source stepping continuation method during harmonic spectra and dc analyses type String the voltage source type magi mag2 Real s 0 0 magnitude of fundamental and harmonics in harmonic analysis As many harmonics as necessary can be used that is the number of harmonics is not upper limited phi ph2 Real s 0 0 phase of fundamental and harmonics in harmonic analysis As many harmonics as necessary can be used that is the number of harmonics is not upper limited mag Real 0 0 signal magnitude in small signal analysis AC analysis phase Real 0 0 signal phase in small signal analysis AC analysis pacmag Real 0 0 magnitude of the source in small signal periodic analysis pac analysis pacphase Real 0 0 phase of the source in small signal periodic analysis pac analysis magit Real 0 0 magnitude of first fundamental in spectral analysis mag2t Real 0 0 magnitude of second tone generated in the spectral analysis freq2t List the list of indices that defines the frequency of the second tone as a linear combination of the two fun
48. phases of the envelope analysis or of the shooting method acntrl Select the accuracy control method 1 energy balance This method computes the electrical energy variation of each capacitor and inductor along each integration interval and the corresponding electrical work delivered by the resistive one port elements to which they are connected The electrical work and energy are compared and the integration time step is chosen so that to keep their difference below the user defined threshold this is the default accuracy mechanism 2 energy balance This method is identical to the previous one but the difference between the electrical work and energy is compared to the total energy stored into the corresponding capacitor inductor The integration time step is chosen to keep variation with respect to total stored energy below the user defined threshold In general this method given less accurate results with respect to the previous one but the analysis runs faster cmin A parasitic capacitor of value lt cmin gt is added between each node user given nodes of the circuit and ground This is a convergence aid for algorithms If it set to 0 its effects are turned off delay Real the time interval along which the time domain analyses transient analyses are performed These transient analyses are carried out by starting from the given initial conditions and compute the solutions at the end of each simulation interval The solutions at the end of
49. resistor instance overwrites the value given in the model type Resistor types 1 linear resistor 2 piece wise linear resistor specified by giving the voltage current corners of its charac teristic see lt v gt and lt i gt parameters for more details i Current value at a corner of the piece wise linear characteristic This parameter has meaning only if the resistor is a piece wise linear one v Voltage value at a corner of the piece wise linear characteristic This parameter has meaning only if the resistor is a piece wise linear one interp The number of samples for characteristic interpolation This parameter has meaning only for piece wise linear resistor types It acts at the corners of the pwl characteristic by smoothing it through the adoption of B spline It is suggested to check if the resulting characteristic is adequate 4 2 2 4 2 Basic one port device 43 scalev Real 1 0 scale voltages of a pwl characteristic scalei Real 1 0 scale currents of the pwl characteristic swapiv Boolean false when set the two columns of the file describing the characteristic of a pwl resistor are swaped Recall that as default the first column represents voltage values and the second one current values pwlfile String the name of the file containig the v i values specifing the characteristic of a pwl resistor Instead of describing the characteristic through the lt v gt and lt i gt parameters it can be read more conve
50. results are referred to the frequency of the relative sweep refh Integer 1 the harmonic to which a relative sweep refers to freq Real the frequency value in single frequency analysis print Boolean false print the result in single frequency analysis harms Real 10 the number of harmonics of the large signal spectra of the unperturbed circuit that are considered in the analysis samples Integer the number of time points employed to sample the large signal steady state solution and to define the size of the related variational model When not specified it is derived from the number of harmonics defined by the harms parameter as 4 harms decomp Boolean false when true the resulting component of the spectra along the ul t eigenfunction and those obtained by difference with it are computed These components have meaning only for autonomous circuits The component along the u1 t eigenfunction is defined as phase modulation and that obtained by subtracting this component from the complete result is defined as amplitude modulation tmax Real maximum integration time step allowed during the transient analysis related to the lt pac gt analysis Before performing a lt pac gt analysis the circuit is linearised along one working period The solution along this working period is determined by a transient analysis The lt tmax gt parameter upper limits the maximum integration time step tha
51. starting from 100 times the value of the lt gground gt parameter and increasing it till the Newton algorithm begins to converge initgmin Real this parameter has effect only when the gmin continuation method is started The initial value of the parasitic conductance is set to the value specified by this parameter instead of starting from 100 times the value of the lt gmin gt parameter and increasing it till the Newton algorithm begins to converge save String the path of the file into which the solution is saved bz2 format is adopted The solution stored in this file can be later loaded by the simulator and used as initial guess to the Newton algorithm load String the path of the file from which a previously saved solution is read The file format can be both ascii and bz2 and it is automatically determined during loading This solution is used as initial guess to the Newton algorithm mem List results that have to be saved in memory These results can be later accessed from the body of the lt control gt analysis through the lt get gt function Note that the saving of results in memory can require and occupy a large amount of memory multiple Boolean false multiple solutions of the electrical circuit are possibly searched Note that the adopted approach does not ensure to find all the solutions of the circuit To find possible solutions some peculiar independent voltage sources are inserted in the circuit The magnitu
52. suitable maximum value that depeneds on the power supply of the oscillator During this sweep the current through the voltage probe or the branch voltage across the current 4 9 4 142 Chapter 4 Reference probe is monitored If the real part or the imaginary part of these electrical variables crosses the zero value the sweep is interrupted and the solution is refined to find the exact amplitude of the probe that cases the branch current of the voltage probe or the branch voltage across the current probe to be 0 This solution is then continued by sweeping the amplitude or the frequency of the voltage probe untill the modulus of the branch current of the voltage probe or the branch voltage across the current probe is 0 When this happens a solution of the steady state problem has been found Note that the insertion of the voltage probe can be automatic that is the method automatically determines the nodes to which connect the voltage probe The newton solver is based on the insertion of a voltage probe or a current probe in the circuit It adopts the Newton iterative method to find the 0 value of the branch current of the voltage probe or of the branch voltage of the current probe The unkowns are the magnitude and the frequency of the probe In general convergence is achieved if the Newton method is started from a suitable initial condition It is suggested to perform a pole zero analysis and to employ the frequency of the unstable pole as i
53. temperature exponent pt Real 3 0 saturation current temperature exponent fc Forward bias depletion capacitance threshold rl Real 1 0e12 leakage resistance connected in parallel to the diode Note that the default gmin conductance is also connected in parallel bv Reverse breakdown voltage 4 4 2 58 Chapter 4 Reference ibv Real 0 001 current at breakdown voltage imax Real 1 0 the maximum current through the diode above which its electrical characteristic is linearized It is also assumed that above this current the diode is melting tnom Parameter measurement temperature trs1 Real 0 0 series resistance first order temperature coefficient trs Real 0 0 series resistance first order temperature coefficient trs2 Real 0 0 series resistance second order temperature coefficient kf Real 0 0 flicker noise coefficient af Real 1 0 flicker noise exponent tlev DC temperature selector tlevc AC temperature selector tvj Temperature coefficient of vj parameter tvjsw Temperature coefficient of vjsw parameter tcjo Temperature coefficient of cjo parameter tcjsw Temperature coefficient of cjsw parameter pta Junction potential temperature coefficient ptp Sidewall junction potential temperature coefficient cta Junction capacitance temperature coefficient ctp Sidewall junction capacitance temperature coefficient level Model selector ns Sidewall emission coefficient Bipolar Junction Transis
54. the cluster and bound to the mosfet instance ELEMENT DEFINITION Name drain gate source bulk ModelName lt parameter value gt ELEMENT PARAMETERS w Channel width 1 Channel length ad Drain region area as Source region area pd Drain perimeter 4 4 semiconductor device 67 ps Source perimeter nrd Drain diffusion sheet resistance nrs Source diffusion sheet resistance temp Device junction temperature m Integer 1 number of devices connected in parallel sa Distance between OD edge to poly of one side sb Distance between OD edge to poly of one other side noisetype Integer 7 types of generated noise possible choices are 0x0 noiseless BSIM3 0x1 thermal noise 0x2 shot noise 0x4 flicker noise The value of this parameter can be the result of the bitwise or of types of noise sources MODEL DEFINITION model ModelName bsim3 lt parameter value gt MODEL PARAMETERS ijth Real 1 0e 1 diode maximum current above which it is considered as melting and its characteristic is linearised to aid numerical methods to converge imax Real 1 0e 1 diode maximum current above which it is considered as melting and its characteristic is linearised to aid numerical methods to converge vbox Real 1 0e9 tox oxide breakdown voltage capmod Capacitance model selector It is an integer number between 0 and 3 default 3 mobmod Mobility model selector noimod Integer 1 noise model selector Possible values are 1
55. the samanskii method At each iteration of the shooting method the sensitivity matrix of electrical variables at the end of the shooting interval with respect to the initial conditions at the beginning of the interval is computed Even though it is a by product of the transient analysis at each time point a matrix by matrix product is needed and it costs CPU time mainly when a large number of electrical variables are involved At each iteration of the shooting method the norm of the residue is monitored when it monotonically decreases of a relative factor lower than 0 9 the computation of the sensitivity matrix is suspended at each subsequent iteration of the shooting method the same previous sensitivity matrix is adopted This leads to an approximate Newton iteration 4 8 3 132 Chapter 4 Reference but it saves CPU time even though in general a larger number of transient analyses are needed damping Real 1 0 the initial value of the damping factor employed by the Newton algorithm to update the new initial conditions Normally at the beginning the Newton iterations are undamped for particular circuit it can be useful to start with damped iterations in order to aid and speed up convergence statevars It is a list of names of device instances for example capacitors and inductors that specify state variables These state variables are monitored by the procedure that determines the working period of the autonomous circuit If this list is
56. the voltage source is output only if parameter print is set to true Finally some execution time statistics are shown The annotate option All analysis admit the option annotate that controls the amount of output generated This parameter is strongly context dependent and allows the user in many cases to output information not in general available in most commercial simulators The general format of the annotate option is a decimal value from 0 to 7 that correspond to increasing level of verbosity plus several bitwise values that act as flags and can be added These flags can be entered as hex constants and are used to enable output of context dependent information As an example modify the 4th row of Simple_001 pan so that it reads DcO dc print yes annotate 3 0x40 0x80 in this case the standard verbosity level 3 is chosen but in addition value 0x40 the simulator is instructed to print the MNA matrix to standard output and to the log file at each Newton iteration step finally value 0x80 the MNA matrix is dumped to file DcO m A large amount of information is printed in our case the last lines of the log file are Logfile 2 2 3 Logfile Simple_001 1log with annotate 3 0x40 0x80 Tescani matrix dump at iteration lt 2 gt MATRIX SUMMARY Matrix is 5 x 5 and real Matrix before decomposition 1 2 3 4 5 1 0 001 0 001 nen aes 1 2 0 001 0 2 1e 12 ses 3 0 2 0 011 0 01 4 T 0 01 0 0
57. these parameters as the input ports This applies only if the VCVS is not modelled by an external macromodel In this case gains are specified by the macromodel gain Real 0 0 gain of each driving port of the VCVS since there can be more than one input port there must be as much of these parameters as the input ports This applies only if the 4 3 3 4 3 Basic two port device 5 VCVS is not modelled by an external macromodel In this case gains are specified by the macromodel delay Time delay max Real maximum allowed value of the output voltage min Real minimum allowed value of the output voltage func The function that describes the characteristic of the VCVS digital Boolean false tells that this is a digital macromodel The term digital is improperly used in fact it is considered as digital a vcvs whose output makes discrete voltage transitions at specific time instants or at specific values of the driving signals The output values have not to be necessarily two boolean This type of vcvs are evaluated only once at each time instant of the transient analysis when the Newton algorithm has reached convergence If the digital vcvs does not change its output value it is considered that the Newton algorithm has definitely converged in the other case the vcvs output is updated and the time step is rejected Please have a look of the lt trtime gt options for further information stvar Integer 0 the num
58. toff There can be several of these time instants period Real switching period it is used only in time domain analyses hindex Integer 1 index of the harmonic at which the switch is operated In spectral analysis the index is referred to the first fundamental MODEL DEFINITION model ModelName switch lt parameter value gt MODEL PARAMETERS ron Real 0 1m resistance of the switch when it is on roff Real 1Meg resistance of the switch when it is off ts Real 0 001 this parameter set the relative time interval with respect to the switching period assumed normalised to 1 during which the switch commutes from the off to the on position or vice versa Being a relative value it must be in the interval 0 1 Transmission Line A transmission line that includes dielectric and conductor loss The conductor loss includes skin effect and does not assume the conductor is infinitely thick All instance parameters are modifiable ZO is default The currents into both ports are calculated Currents are defined positive if they flow into the inner conductor through the transmission line and then out of the outer conductor Only the odd mode is modeled so only the voltage difference across each port is important not the absolute voltage of each terminal To model both even and odd modes it is necessary to combine two lines The line that models the odd mode will have both ports floating the line modeling the even mode will hav
59. ub1 puc Cross term dependence of uc puci Cross term dependence of uc1 pu0 Cross term dependence of u0 pute Cross term dependence of ute pvoff Cross term dependence of voff pelm Cross term dependence of elm pdelta Cross term dependence of delta prdsw Cross term dependence of rdsw pprwg Cross term dependence of prwg pprwb Cross term dependence of prwb pprt Cross term dependence of prt petaO Cross term dependence of eta0 petab Cross term dependence of etab ppclm Cross term dependence of pclm ppdiblc1 Cross term dependence of pdiblc1 ppdiblc2 Cross term dependence of pdiblc2 ppdiblcb Cross term dependence of pdiblcb ppscbei Cross term dependence of pscbel ppscbe2 Cross term dependence of pscbe2 ppvag Cross term dependence of pvag pwr Cross term dependence of wr pdwg Cross term dependence of dwg pdwb Cross term dependence of dwb pbO Cross term dependence of b0 pbi Cross term dependence of b1 pcgsl Cross term dependence of cgsl pcgdl Cross term dependence of cgdl pckappa Cross term dependence of ckappa pef Cross term dependence of cf pclc Cross term dependence of clc pcle Cross term dependence of cle palphaO Cross term dependence of alphaO palphai Cross term dependence of alphal pbeta0O Cross term dependence of beta0 pvfbcv Cross term dependence of vfbcv pvfb Cross term dependence of vfb pacde Cross term dependence of acde pmoin Cross term dependence of moin pnoff Cross term dependence of noff pvoffcv Cross term dependence of
60. view Please refer to the cited paper and its bibliography for technical details concerning the used algorithms A simple Analog Digital oscillator The first circuit we will consider is a very simple Pulse Energy Restore Oscillator PERO BBS11 Biz 11 The schematic of the PERO oscillator is shown in Fig 3 1 It is composed by Digital Figure 3 1 The schematic of the PERO oscillator C 1F L 1H R 10Q R 100MQ R3 50Q E 1V ie Se e e 36 Chapter 3 AMS simulation with PAN a linear analog part and a digital Finite State Machine that recharges the energy in the LC tank A zero crossing detector zcd is used as input to a 2 bit digital counter cnt that closes switch S when both its bits are high The analog netlist is written in standard PAN form and is shown in Netlist 3 1 1 Netlist 3 1 1 The PERO oscillator analog part File pero pan ground electrical gnd parameters VDD 1 verilog_include comb v Tran tran tstop 300 uic yes ShO shooting period 25 autonomous yes restart no method 2 order 2 solver 2 minper 19 floquet yes Pnoise pnoise start 10u stop 100m dec 4000 onodes x sweeptype 0 ri x gnd resistor r 10 cl x gnd capacitor c 1 ic 1 11 x gnd inductor 1 1 SW x vdd y gnd SW vr vdd gnd vsource vdc VDD model SW vswitch ron 5 roff 100M von 0 6 VDD voff 0 4 VDD A2d1 x gnd a2d dignet cnt X vl 0 vh 0 D2al y gnd d2a dignet cnt Y vl 0 vh VDD D2a2 z gnd d2a dignet c
61. voltages lt input gt double the array i of the port currents lt output gt double the G matrix lt output gt double the Ai matrix lt output gt double the R vector of known terms lt input gt double the value of simulation time It has meaning only in time domain analyses The possible routine handling parameters of the macromodel is called with the arguments lt input gt int the index that uniquely identifies the instance of the N port lt input gt char parameter name lt input gt double parameter value It must return 1 upon successful execution and 0 otherwise ELEMENT DEFINITION Name port portl lt a lot of terminals gt ModelName lt parameter value gt ELEMENT PARAMETERS funci The functions that describe the implicit characteristic of the NPORT There must be one describing function for each port of the NPORT Ports are formed by pairs of 98 Chapter 4 Reference terminals each port is identified in the expression by the p character followed by the port number Ports are numbered by starting from 1 For example a linear resistor of 2 ohm can be decribed as Nport1 pos neg nport funcl v p1 2 i p1 sensedev String the name of the instance of the sensing device The current through this device is used to control the NPORT If the instance of the NPORT belongs to a subcircuit the sensing device is searched locally This means that is a local and global device
62. waveform It must be expresse as an absolute time interval and not as a relative number with respect to the period of the pulse waveform period Real 0 0 the period of the pulse and pwl generator If not specified for a pwl generator then the waveform is continued above the last time point by maintaining the last voltage value it is continues as a constant waveform 4 3 4 3 1 4 3 Basic two port device 49 Basic two port device Voltage Controlled Current Source Voltage controlled current source is a device with at least 2 terminals current exits the source terminal and enters the sink terminal The number of terminals must be even those above the first 2 are the controlling terminals There is a lt gm gt parameter for each controlling pair that linearly relates the output current to the voltage across the corresponding input port External macromodels are supported The routine describing the macromodel is called with the following ordered sequence of arguments lt input gt int an index that uniquely identifies the instance of the VCCS This index starts from 0 and it is incremented by 1 each time a new instance is found in the netlist lt input gt int number of ports of this VCCS lt input gt double the array of the port voltages lt input gt int number of state variable lt input gt double the two vectors of state variable values at the current n time point index 0 and at the previous n 1 time point index 1
63. 00 0 Celsius maximum allowed device temperature vmax Real maximum voltage allowed during iterations of the Newton algorithm vmin Real minimum voltage allowed during iterations of the Newton algorithm vmaxguess Real 0 0 maximum voltage allowed during evaluation of the initial guess of the Newton algorithm If a node voltage is above this value it is set to the vmax value By setting vmax to 0 and vmin to 0 the Newton algorithm starts with a 0 initial guess and thus the evaluation of the starting value for the Newton algorithm has no effect vminguess Real 0 0 minimum voltage allowed during evaluation of the initial guess of the Newton algorithm If a node voltage is below this value it is set to the vmin value 108 Chapter 4 Reference ireltol Real 1 0e 6 relative current convergence criterion iabstol Real 1 0e 12 absolute current convergence criterion vreltol Real 1 0e 6 relative voltage convergence criterion vabstol Real 1 0e 6 absolute voltage convergence criterion ic List it is a list specifying the initial value of node voltages and branch currents employed as a guess for the Newton iterative method These values override those automatically computed by the lt dc gt analysis as initial guess A dc analysis is initially done where the ic conditions force the Newton method When convergence has been achieved the Newton method runs free to find the operating point of the circuit with no more c
64. 1 5 1 Largest element in matrix 1 Smallest element in matrix 1e 12 Largest pivot element 1 Smallest pivot element 0 01 Number of elements 12 Density 48 00 Number of fill ins 0 16 Chapter 2 A simple PAN walk through SSsSSso5 RHS vector Beginning source vector 4 25689263434929e 19 3 02872507961481e 19 1 71973044932689e 19 8 58284610952035e 20 oP WN FB Delta solution vector Beginning source vector 1 0 2 1 5219724074667e 18 3 1 1243919682759e 15 4 1 13297481427212e 15 5 4 24167291027462e 19 Current solution of lt Dc0 gt DC analysis V n1 2 0000 V V n2 10 050 mV V n3 2 0201 V V n4 1 9201 V I vi i 2 0101 mA Matrix element number 12 Fill ins number 0 Sparsity percentage 52 Reordering number Exchaged rows cols Subcircuit solutions oor 2 2 2 DC sweep and device models A common use of the DC analysis is not limited to the computation of one single operating point but to the computation of operating points as the value of one circuit parameter is varied This analysis mode is known as sweep and most simulators provide a DC sweep analysis In PAN DC sweep can be performed through a number of different options that control the parameter or option to be swept and the sampling of the values to be used for the sweep i e linear logarithmic Instance parameters i e resistance value of a resistor value of an independ
65. 3 narrate parser actions bit 4 print some values of some device instances bit 5 print some info about digital nets warn Boolean true give warning messages maxwarns Integer 10 maximum number of given warnings The other ones are omitted in this case the total number of warnings given and not given is displayed at the end of the analysis maxnotices Integer 10 maximum number of given notices The other ones are omitted in this case the total number of notices given and not given is displayed at the end of the analysis topcheck Integer 1 turns on checking of circuit topology for degeneracies Valid values are 0 disable any topology check 1 a concise report about topology errors and degenerations 2 a more verbose reporting ignoreshorts Boolean 0 shorted devices are ignored that is they do not contribute to the analyses On the other hand if set false they are considered in the analysis This is usefull for example when there are mosfets with drain and source terminals connected together or when there are multi port controlled sources with one shorted input port In this case the entire controlled source will be ignored if the ignoreshorts option is true unusedmod Boolean false print the name of the models that are defined in the netlist but not referred to by any instance ignoreundefinstpars Boolean 0 undefined parameters of instances are ignored This may be usefull to allow the reading of netlists of f
66. 8 Advanced time domain analyses 123 4 8 1 Time domain analysis with envelope algorithm saaa a 123 4 8 2 Steady state computation through shooting method 126 4 8 3 Steady state computation through the multi tone shooting method 132 4 8 4 Periodic Small Signal Andlysis 0 aaa 134 4 8 5 Noise analysis of a periodic Circuit 6 es 135 49 Advanced frequency domain analyses 137 4 9 1 Pole Zero Analysis 2 eee ee tenes 137 4 9 2 Periodic Steady State Analysis o a aaa 137 4 9 3 Almost Periodic Steady State Analysis 0 0 es 140 4 9 4 Small Signal S Parameter Analysis 0 0 aaa 142 4 9 5 Noise analysis of a periodic Circuit uaaa aaa 143 eji o aA E E E E E E E T 145 Books 145 Articles 145 Technical Reports 146 Sind OLY ITN Ve SO E rT I i P P I Nes oF ait Hitt Ny Oat Ranie gt th Va fy Reve Wire Be Ag itl POVES LJ ee Free Cer he Set mu lest KOZA PAR Uli ee Bar EIA PAS Se pa a ea Ea CAH S E bal es IVE d Mpa De eR es sail n EAEN wi t We tats 7 Sa EIA ENIKI Ehh y al er atte ap Sa roe en i SEOs esie kz ninsa 3 EOR ite Te PAARI i z mhe Orta i i DUS T zt ei y AES IN Ste AN nee A r S Se Ne A e aE gah lE aN R Oa N SR ane oe en UE RE SAN MEERN A Ms iii ete Ty 3 elites R MEEA ENEON S EEE p a oel OR K r H IN IEI olig o FS lt ali We aes s lu 1 Preface hee aA REALS ji OPM FA AH Sh DoT UONI 1 x Coie 4 22 FO lo
67. AIL SY Ufo x i ars t P D ke 4 gt A gt UB qi s At Y i i gt we BI f A a Dre 9A e E i S on ts MET Hones PNA ec ay if I a fol I Tore ROU FANII Hiii AENA REEE E E AONE A ERANO AAEE y a a e aa Oa TOE zil MoE R ty yy nga Tr a Ss 2 peo F z in all lt N ail j as wees la gt p mv uf TET r ESIIN a reng e POPU R SARE SIE NG SILLS Sagas SST aban ta gt PN L A ie e E E E T 7 2 A simple PAN walk through 0 200ceee cece eens 9 2 1 Overview 9 2 1 1 Where and How downloading and installing s saaa 9 2 1 2 Online help and quick reference uuaa aaa aa 9 ZS Netist FIGS sese isa bance eed ace ia bd bette a thn aa deb a cache he 10 2 14 Models and devices ci ee 11 2 1 5 Global OPTIONS 6 seue ea d ae OLE E a OR ee ER ke Oa 11 2ko Output format s sire neeaae a e Sedan aka AE E E a 11 2 2 The DC analysis 12 2 2 1 A very simple example 0 cc eee eee 12 2 22 DC sweep and device models 0 00 ee 16 2 2 3 The alter PSCUGO ANAIYSIS 0 eee eee 17 2 2 4 Multiple DC SWEEPS 1 eee eee eens 19 2 2 5 Multiple DC operating points and simulator directives 0 0 0 20 2 3 The time domain analysis tran 23 2 3 1 A Vander Pol oscillator parameters expressions and the poly device 23 2 3 2 A simple linear LC oscillator and frequency warping effects 26 2 4 Envelope time domain analysis envelope 27 2 4 1 The Bubba oscillator using Schem
68. Bre Belew i Ui ental N et Sa 3 SOOT UO FICS E EATE BOL fesse US SRL OY eens eens eo O pre eH Sy Pope Go ONA This book represents a hands on introduction to the PAN circuit simulator Since our goal was to provide a fast track to potential users it has been formatted as a walkthrough that starts from the simplest concepts and evolves towards more complex methods and analyses PAN is an accademic simulator it has many powerful features it is free but obviously most certainly has bugs of which we are not aware We are quite eager to correct them and also to implement new user requested features but users may not expect the same kind of excellent support that comes with commercial simulators PAN has also a few features that to our knowledge are not implemented in currently available commercial simulators The most notable of these is the capability to perform steady state and noise analyses on circuits that are built using a mixture of analog digital and behavioural blocks A simple example of the use of these algorithms is shown in a separate chapter IMPORTANT NOTE this book is work in progress In its current status it can be used to start using PAN in a simple and direct way we intend to expand it with many details and considerations on the algorithms that currently are missing If you are interested please check on the PAN distribution site for new versions F See Oe Malas Sion ORES en ee on
69. C junction exponent mc B C junction exponent xcjc Fraction of B C capacitance tied to internal base node cdis Fraction of B C capacitance tied to internal base node cjs Real 0 0 c substrate zero bias junction capacitance ccs Real 0 0 c substrate zero bias junction capacitance 4 4 3 60 Chapter 4 Reference vjs C Substrate built in junction potential mjs Substrate junction exponent ms Substrate junction exponent fc Junction capacitor forward bias threshold tf Ideal forward transit time xtf Coefficient for bias dependence of tf vtf Voltage describing Vbc dependence of tf itf High current parameter for effect on tf jtf High current parameter for effect on tf ptf Excess phase at freq 1 0 tf 2pi Hz tr Real 0 0 ideal reverse transit time xtb Beta temperature exponent tb Beta temperature exponent eg Band gap xti Temperature exponent for effect on Js pt Temperature exponent for effect on Js tnom Parameter measurement temperature kf Real 0 0 flicker noise coefficient af Real 1 0 flicker noise exponent Junction Field Effect Transistor The JFET model is derived from the FET model of Shichman and Hodges JFETs require the use of a model statement ELEMENT DEFINITION Name drain gate source ModelName lt parameter value gt ELEMENT PARAMETERS area Junction area in square meters region DC operating region O off 1 on 2 rev 3 ohmic noisetype Integer 7 types of generated noise
70. In other words the array can be indexed as array 0 StateVarNum current time point and array 1 StateVarNum previous time point lt output gt double the array of the transconductances the meaning is the same as that of the lt gm gt parameter lt output gt double the value of the current flowing through the output port lt input gt double the value of simulation time It has meaning only in time domain analyses The possible routine handling parameters of the macromodel is called with the arguments lt input gt int an index that uniquely identifies the instance of the VCCS lt input gt char parameter name lt input gt double parameter value It must return 1 upon successful execution and 0 otherwise ELEMENT DEFINITION Name sink source lt a lot of terminals gt ModelName lt parameter value gt ELEMENT PARAMETERS gaint gain2 Real 0 0 gain of each driving port of the VCCS since there can be more than one input port there must be as much of these parameters as the input ports This applies only if the VCCS is not modelled by an external macromodel In this case gains are specified by the macromodel gm Real transconductance This parameter is equivalent to lt gain1 gt delay Real 0 0 time delay func The function that describes the characteristic of the VCCS offset Real 0 the offset current generated by the VCCS This current is added to that due to the driving signals
71. MOS2 model adopted for channel thermal noise and fliker noise 2 BSIM3 model adopted for channel thermal noise and fliker noise 3 MOS2 model adopted for channel thermal noise and BSIM3 for fliker noise 4 BSIM3 model adoted for channel thermal noise and MOS2 for fliker noise quiet Boolean false when true warning messages about parameters whose values are out of the allowed range are given binunit Bin unit selector tox Gate oxide thickness in meters nqsmod Non quasi static model selector It is a boolean flag default false nmos Nmos type of mosfet toxm Gate oxide thickness used in extraction cdsc Drain Source and channel coupling capacitance cdscb Body bias dependence of cdsc cdscd Drain bias dependence of cdsc cit Interface state capacitance nfactor Subthreshold swing Coefficient xj Junction depth in meters vsat Saturation velocity at tnom at Temperature coefficient of vsat a0 Non uniform depletion width effect coefficient ags Gate bias coefficient of Abulk a1 Non saturation effect coefficient a2 Non saturation effect coefficient keta Body bias coefficient of non uniform depletion width effect nsub Substrate doping concentration nch Channel doping concentration 68 Chapter 4 Reference ngate Poly gate doping concentration gammai Vth body coefficient gamma2 Vth body coefficient vbx Vth transition body Voltage vbm Maximum body voltage pmos Pmos type of mosfet xt Doping depth k1 Bulk effect co
72. ON Name pos neg a2d lt parameter value gt ELEMENT PARAMETERS 4 5 3 96 Chapter 4 Reference dignet String the hierarchy path inside the design that identifies the net to which the converter is connected For example Mod U1 e specifies the net e of instance U1 that is in turn instantiated in the Mod root module vl Real the voltage threshold that defines the low logic state The analog voltage is coded in the corresponding binary value All the bits of the digital coding are at the logic level 1 for analog values greater than lt vh gt and all the bits are at the logic level 0 for analog values less than lt vl gt When the digital net is composed of only 1 bit the value of this parameter is set equal to that of the lt vh gt one If the values of lt vh gt and lt vl gt parameters differ the average value is taken vh Real the voltage threshold that defines the high logic state In case of a digital net variable composed of more than 1 bit the analog voltage is coded in the corresponding binary value All the bits of the digital coding are at the logic level 1 for analog values greater than lt vh gt and all the bits are at the logic level 0 for analog values less than lt vI gt When the digital net is composed of only 1 bit the value of this parameter is set equal to that of the lt vl gt one If the values of lt vh gt and lt vl gt parameters differ the average value is taken vt R
73. Sweep limit span step Step size linear sweep lin Number of points linear sweep dec Points per decade log Number of points log sweep sweeptype Integer 2 type of frequency sweep Possible ones are 0 absolute sweep 1 relative sweep 2 offset sweep When absolute sweep is chosen the frequency is swept in the interval defined exclusively by the sweep When the relative sweep is chosen the frequency is swept in the interval specified by the sweep with respect to the fundamental frequency corresponding to the working period determined by the large signal steady state analysis In other words the frequency of the sweep add to the oscillator fundamental When the offset sweep is chosen the frequency is swept as in the relative sweep but the results are referred to the frequency of the sweep This parameter acts also in single frequency analysis define by the lt freq gt parameter refh Integer 1 the harmonic to which a relative sweep refers to freq Real the frequency value in single frequency analysis This parameter must not be specified contemporary to the frequency sweeping command onodes List the list containing the pair of nodes the ground node can be omitted at which the noise is computed The format is lt onodes node1 node2 gt harms Real 7 the number of harmonics of the large signal spectra of the unperturbed circuit that are considered in this analysis samples Integer the nu
74. The PAN book Federico Bizzarrif SS Angelo Brambill Giancarlo Storti Gajan M i 4 version 2014 3 a Copyright 2014 Giancarlo Storti Gajani PUBLISHED BY PUBLISHER BOOK WEBSITE COM Licensed under the Creative Commons Attribution NonCommercial 3 0 Unported License the License You may not use this file except in compliance with the License You may ob tain a copy of the License at http creativecommons org licenses by nc 3 0 Unless required by applicable law or agreed to in writing software distributed under the License is distributed on an AS IS BASIS WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND either express or implied See the License for the specific language governing permissions and limitations under the License First printing March 2015 P r I PN Ess Rage RA nf EE TT Soe i hi Mi x BT Vea Sift eae AV RU Mat Sen toss E Mt eM U RS RC MS 2 M gt gt 7 eet hot G PUO ETSEN Ss NaH SE IMIT E i k EE eR EE ek es EN Ns ea oe coe SFOS Wea Nh E ea Sc ce ah ne Tee aA oa GS ea Ra FS CaS EAA IIO HR a DOAI RES MIEREA N ere IVS Satara h SOES SEM Wes Sas ii SERNER UES PEA VER ji A 24 redteeld SEES Se nfi A aS E 45 ee hg r EEEN et N Glo Seat LSet MN bac SLAY SES OS SEES WEARS IS Blast alll ae net sty AT MEE 7 ii f Ry 9 t 4 K IIE gG gt ee OA A y 1 t 4 J IG A re teas r Ia Comes SUN OTE EN re A i EAR
75. Width dependence of ub1 wuc Width dependence of uc wuci Width dependence of ucl wud Width dependence of u0 wute Width dependence of ute wvoff Width dependence of voff welm Width dependence of elm wdelta Width dependence of delta wrdsw Width dependence of rdsw wprwg Width dependence of prwg wprwb Width dependence of prwb wprt Width dependence of prt weta0 Width dependence of eta0 wetab Width dependence of etab wpclm Width dependence of pclm wpdiblc1 Width dependence of pdiblc1 wpdiblc2 Width dependence of pdiblc2 wpdiblcb Width dependence of pdiblcb wpscbei1 Width dependence of pscbel wpscbe2 Width dependence of pscbe2 wpvag Width dependence of pvag wwr Width dependence of wr wdwg Width dependence of dwg wdwb Width dependence of dwb wbO Width dependence of b0 wb1 Width dependence of b1 wcgsl Width dependence of cgsl wcgdl Width dependence of cgdl wckappa Width dependence of ckappa wef Width dependence of cf wclc Width dependence of clc 74 Chapter 4 Reference wcle Width dependence of cle walphaO Width dependence of alpha0 walphai Width dependence of alphal wbeta0 Width dependence of beta0 wvfbcv Width dependence of vfbcv wvfb Width dependence of vfb wacde Width dependence of acde wmoin Width dependence of moin wnoff Width dependence of noff wvoffcv Width dependence of voffcv pcedsc Cross term dependence of cdsc pcedscb Cross term dependence of cdscb pcdscd Cross term dependence of cdscd pcit Cross term dep
76. able has converged if dX n lt reltol Y n abstol The ntoltype parameter defines three different ways to compute Y n Their meanings are listed in the following 1 global Y n is defined as the maximum over the time interval till the current time point of the maximum value assumed by the X n electrical variables where n varies from 1 to N being N the number of circuit variables Note that in this case Y n is a unique value adopted for all the dX n variable variations 2 local Y n is defined as the maximum absolute value of X n in the time interval till the current time point Note that in this case Y n is different for each electrical variable 3 point local Y n X n I The convergence check tightens as ntoltype increases acntrl Integer 1 select the accuracy control method skip cmin 1 relative energy balance this method computes the electrical energy variation of each capacitor and inductor along each integration interval and the corresponding electrical work delivered by the resistive one port elements to which they are connected The electrical work and energy are compared and the integration time step is chosen in order to keep their difference below the user defined threshold 2 absolute energy balance This method is identical to the previous one but the difference between the electrical work and energy is compared to the t
77. age value it is continues as a constant waveform ev The voltage of the expotential source tdi Delay time in the application of the voltage rise by the exponential voltage source tau1 Rise time constant of the exponential voltage source td2 Delay time in the application of the voltage fall by the exponential voltage source tau2 Fall time constant of the exponential voltage source fmfreq Real 0 0 frequency of the phase modulating signal Phase modulation is active only in time domain analyses such as tran envelope shooting and mtshooting fmmag Real 0 0 maximum value of the phase modulating signal Phase modulation is active only in time domain analyses such as tran envelope shooting and mtshooting The frequency modulation acts on the sinusoidal signal generation that must be thus contestually specified as part of the same voltage source fmphase Real 0 0 phase of the phase modulating signal Phase modulation is active only in time domain analyses such as tran envelope shooting and mtshooting Independent Current Source Positive current exits the source node and enters the sink node The only meaningful parameter in a DC analysis is idc In an AC analysis only mag and phase are meaningful In a harmonic analysis idc magl mag2 phasel phase2 etc are used In a spectral analysis idc magIt phaselt mag2t and phase2t are used In the time domain analyses such as transient shooting envelope multishooting the piece
78. ain analyses 139 choice of lt thresh gt can improve the speed of the harmonic balance method On the other hand it introduces also inaccuracy and even it can prevent convergence that is it causes a drop in the effectivenes and efficiency of the harminic balance method Therefore it must be carefully used saman Real 2 0 samanskii constant At each iteration of the Newton algorithm the norm of the residue right hand side vector is computed and compared to that computed at the previous iteration If the ratio between the two norms is less than the value of the samanskii constant plus one then the Jacobian is recomputed In general Jacobian recomputation speeds up convergence autonomous Integer 0 solve an autonomous circuit The following different methods to find a steady state solution can be used 1 continuation 2 continuation with no hybrid newton acceleration 4 hybrid newton 5 conventional The continuation solver is bases on the insertion of a probe in the circuit It is composed of different phases In the first phase the magnitude of the probe is swept starting from 0 till a suitable maximum value that depeneds on the power supply of the oscillator During this sweep the current through the voltage probe or the branch voltage across the current probe is monitored If the real part or the imaginary part of these electrical variables crosses the zero value the sweep is interrupted and the solution is refined to find t
79. al 0 0 constant term at numerator ni Real 0 0 coefficient of first degree term of the numerator n2 Real 0 0 coefficient of second degree term at numerator n3 Real 0 0 coefficient of third degree term at numerator n4 Real 0 0 coefficient of fourth degree term at numerator n5 Real 0 0 coefficient of fifth degree term at numerator n6 Real 0 0 coefficient of sixth degree term at numerator n7 Real 0 0 coefficient of seventh degree term at numerator n8 Real 0 0 coefficient of eighth degree term at numerator dO Real 1 0 constant term at denominator di Coefficient of first degree term at denominator d2 Coefficient of second degree term at denominator d3 Coefficient of third degree term at denominator d4 Coefficient of fourth degree term at denominator d5 Coefficient of fifth degree term at denominator d6 Coefficient of sixth degree term at denominator d7 Coefficient of seventh degree term at denominator d8 Coefficient of eighth degree term at denominator tnom Real 27 nominal temperature at which the polynomial characteristic was determined tcl Real 0 0 coefficient of the linear term in the temperature dependence relation tc2 Real 0 0 coefficient of the quadratic term in the temperature dependence relation noisetype Integer 5 types of generated noise possible choices are 0x0 noiseless element 0x1 thermal noise 0x4 flicker noise The value of this parameter can be the result of the bitwise o
80. al solutions at each iteration of the Newton method 0x20 print MNA matrix and RHS vector at each iteration of the Newton method checkstrange Boolean 1 check if there are some strange behaviours of devices For example it is checked if bulk drain and bulk source parasitic diodes of mosfets are forward biased This check slows down a bit the simulation and can generate a large number of warnings safetydigits Integer 0 how many more digits of the ALU should be preserved by the large conductace transformation maxdevtemp Maximum allowed device temperature Celsius maxiter Integer 10 maximum allowed number of Newton iterations at each time step of the time domain analysis The Newton algorithm is considered as non converging when the iteration number goes above the maxiter value In this case the integration time step is shortened and the Newton algorithm is restarted saman Boolean false turn on off the Newton Samanski method maxinititer Integer 100 maximum allowed number of Newton iterations during the deter mination of the initial conditions from which the transient analysis is started maxtminiter integer 100 maximum allowed number of Newton iterations when the inte gration time step length is equal to the its minimum allowed value specified through the tmin parameter printnodes Boolen false print external internal node names index map This parameter serves mainly for debug purposes savelist It def
81. alk through up meaning that a full tran analysis could be to expensive in terms of time to perform On the other hand if a first short tran is not performed the long envelope simulation that follows does not converge After invoking the simulator a large number of log messages are output The 1200p 1800 Jj x os 560 AAA O 10V 140n a Re aN 47p 15p Ta E 2N5179 S Figure 2 13 The Butler overtone oscillator largest part of the messages come from the envelope analysis and are not interesting in this context The most relevant messages are those relative to the last shooting analysis A small portion of the log file is shown below in 2 5 2 Logfile 2 5 2 A snip of the logfile of the Butler Oscillator butler log Notice given by pan during shooting analysis ShootGE5 The relative error in identifying the Floquet multiplier theoretically equal to 1 is 20 872 u the second multiplier gives a relative error equal to 35 319 u the scaling factor is equal to 652 90 G Floquet multipliers with real part larger than 9 99998e 07 5 3635e 01 9922e 01 3356e 01 3 4287e 02i1 0 83426 3356e 01 3 4287e 02i1 0 83426 9349e 01 0000e 00 9991e 01 4138e 06 O O m e 000a First Floquet exponent 783 27 k These messages are strictly related to the algorithm used to compute the shooting solution Their meaning goes bey
82. ample consider Sweep sweep start 0 stop 1 step 0 1 instance r1 mioparam r begin Tran tran tstop 1m 4 7 4 118 Chapter 4 Reference Shoot shooting autonomous 1 fund 1k restart no end A sweep named lt Sweep gt acts on the parameter lt r gt of the instance lt r1 gt Parameter value is linearly swept in the 0 1 interval by steps of value 0 1 At each point of the sweep the lt Tran gt transient analysis is executed till the lt stop 1m gt time stop After the lt Tran gt analysis the lt Shoot gt shooting analysis is executed starting from the results computed by the previous lt Tran gt analysis In the example above the sweep acts on an instance parameter Consider the example parameters p1 0 p2 0 Sweep sweep start 0 stop 1 step 0 1 param p1 start 0 stop 1 step 0 1 param p2 begin Tran tran tstop 1m Shoot shooting autonomous 1 fund 1k restart no end the p1 and p2 parameters defined through the parameters keyword are varied and the same analyses described above are executed for each value assumed by each parameter Note that this sweep is two dimensional i e the sweep generates a square grid of samples There is no upper limit to the number of sweeps in the same sweep card ANALYSIS DEFINITION Name sweep lt parameter value gt ANALYSIS PARAMETERS annotate Integer default is global annotate level resetsweep Boolean true set back all sweeping parameters at the values they had at the beginning o
83. and the x result is a matrix The computed spectrum is monolateral If the number of samples is even the component at the Nyquist frequency is stored as the last entry of x flushQ flush output fflush x flush the file identified by the x file descriptor fd fopen filename mode open a file for read write accorting to the mode selector Possible modes are those of standard unix The file descriptor is returned in the fd variable fprintf x get result index get simulation results instance and model parameter values The result string identifies the result in the memory waveform data base The syntax is NameOfTheAnalysis Eqn where NameOfTheAnalysis refers to a simulator analysis and Eqn is the name of a circuit node or branch equation If a parameter of an instance or model has to be retrived the syntax is similar to the previous one The name of the analysis has to be substituted with the name of the instance or model the node name has to be substituted with the parameter name If a parameter of an instance does not exist it is considered as a parameter of the corresponding model associated to the instance and the search is repeated The second parameter of get is an integer if not supplied the entire result vector is retrieved If index is equal to 0 the current size of the resulting vector is retrieved If it is a positive integer the first index elements of the result vector
84. annel width ad Drain region area as Source region area pd Drain perimeter ps Source perimeter nrd Drain diffusion sheet resistance nrs Source diffusion sheet resistance region Operation region it is used to compute the guess to star the Newton method during the computation of the DC operating point Allowed values are region 0 off region 1 on in the first case the Vgs voltage is set equal to Vth and Vds 2 Vth in the second case Vgs 0 Default value of parameter is region 1 temp Device junction temperature noisetype Integer 7 types of generated noise possible choices are 0x0 noiseless resistor 0x1 thermal noise 0x4 flicker noise The value of this parameter can be the result of the bitwise or of types of noise sources 64 Chapter 4 Reference MODEL DEFINITION model ModelName mos2 lt parameter value gt MODEL PARAMETERS nmos Nmos type transistor pmos Pmos type transistor type String the mosfet type Available keywords are lt n gt and lt p gt cbd Real 0 0 bulk drain junction capacitance cbs Real 0 0 bulk drain junction capacitance gamma Bulk threshold parameter kp Transconductance parameter phi Surface potential is Bulk junction saturation current pb Bulk junction potential cgso Real 0 0 gate source overlap cap cgdo Real 0 0 gate drain overlap cap cgbo Real 0 0 gate bulk overlap cap cj Bottom junction capacitance per area mj Bottom grading coe
85. annotate 5 tmax Period 100 Envel envelope tstop 250k 0 55 Period method 1 restart no autonomous yes tmax Period 1k period Period ShootGE4 shooting period Period printnodes yes uic yes restart no autonomous yes method 2 order 4 tmax Period 1k floquet yes 3 Circuit qi 0 b e neg Q2N5179 re e neg resistor r 560 rb b n200 resistor r 33 c1 n200 loop capacitor c 15p icon 20 c2 loop 0 capacitor c 47p r2 n200 0 resistor r 2 7k al n200 n300 inductor 1 140n co n300 0 capacitor c 1200p mil n300 neg resistor r 1 8k lo n400 loop inductor 1 420 82496n xtl n400 loop xtal vn O neg vsource vdc 10 vlo e n400 vsource vdc 0 define xtal termi term2 parameters RM 93 co termi term2 capacitor c 4 18p rm termi ni00 resistor r RM cm n100 n200 capacitor c 0 00026p 1m n200 term2 inductor 1 6 76557m end model Q2N5179 bjt npn yes is 69 28a xti 3 eg 1 11 vaf 100 bf 282 1 ne 1 177 ise 69 28E 18 ikf 22 03m xtb 1 5 br 1 176 nc 2 isc 0 rc 4 cjc 893 1f mjc 3017 vjc 75 fc 5 cje 939 8f mje 3453 vje 75 tr 1 588n tf 141 1p itf 27 vtf 10 xtf 30 rb 10 Note that several analyses are performed a first dc analysis is followed by a first relatively short tran that in turn is followed by a relatively long envelope starting from the results of the previous tran and finally by a shooting analysis This oscillator is not easy to simulate since it requires a very large number of periods to power 32 Chapter 2 A simple PAN w
86. aram vdc start 85 0 step 0 01 stop 10 0 instance ri param r start 1 0 step 2 0 stop 11 0 3 Circuit vi ni O0 vsource vdc 2 di ni 0 mydiode ri ni 0O resistor r 1 Models model mydiode diode is 3 142f rs 3 536 n 1 xti 3 eg 1 11 m 0 4186 vj 0 75 fc 0 5 bv 80 ibv 24 573m imax 100 The resulting output is stacked columnwise in output file Simple_004 raw DcO after destacking most wave viewer software can take care of this we have a set of waveforms such as those shown in Fig 2 4 2 2 5 20 Chapter 2 A simple PAN walk through 100 T 1 r 80 f 60f 20 f 20 1 L 1 L 90 70 50 30 10 10 Figure 2 4 Voltage source current vs generator voltage for different values of parallel resistance Netlist 2 2 6 Multiple DC operating points and simulator directives We now present a non standard option of PAN s DC analysis algorithm Even if this option may seem somewhat beyond the scope of a walk through we believe it is sufficiently simple and important to be included here Most non linear circuits can have more than one DC solution These solutions correspond to operating points and in time domain to equilibria PAN implements a heuristic method that can find multiple operating points when they are present Note that the algorithm may not find all solutions in some cases but has proven to work quite nicely in many situations To our knowledge in current commercial or non comm
87. at all the capacitors and inductors whose initial conditions have not been specified are open circuit and short circuit respectively In this case suitable voltage source and capacitor loops orcurrent source and inductor cut sets do not cause any problem period Real 0 0 if the circuit is autonomous it is an estimation of the working period if the circuit is non autonomous it represents the fixed working period of the circuit and it is not computed by the shooting analysis If the initial guess is loaded from a file storing results results by a previous shooting analysis and this parameter is omitted the period is set equal to that of the previous loaded analysis If the initial guess is taken from a previous shooting analysis successfully terminated before the current one and this parameter is omitted the period is set equal to that of the previous analysis fund Real the working frequency of the circuit in case of nonautonomous circuits or an estimation of the working frequency in case of autonomous circuits This estimation is refined till computing the exact working frequency by the shooting analysis freltol Real 1 0e 4 relative tolerance employed by the shooting analysis to check conver gence in determining the working frequency of autonomous circuits tinc Real 1 1 lengthening factor of the time step at convergence tmax Maximum allowed time step length tmin Minimum allowed time step length saman Boolean no turns on off
88. atic Capture and Netlisting tools 27 2 5 Steady state time domain analysis shooting 29 2 5 1 The Butler overtone oscillator 2 aa 30 3 AMS simulation with PAN 000 00 000 e eee 35 3 1 A simple Analog Digital oscillator 35 3 2 The pnoise analysis 38 4 1 4 2 4 2 1 4 2 2 4 2 3 4 2 4 4 2 5 4 3 4 3 1 4 3 2 4 3 3 4 3 4 4 3 5 4 3 6 4 3 7 4 3 8 4 4 4 4 1 4 4 2 4 4 3 4 4 4 4 4 5 4 4 6 4 4 7 4 4 8 4 4 9 4 4 10 4 5 4 5 1 4 5 2 4 5 3 4 5 4 4 5 5 4 5 6 4 5 7 4 5 8 4 5 9 4 5 10 4 6 4 6 1 4 6 2 4 6 3 4 7 4 7 1 4 7 2 4 7 3 4 7 4 KOISO O erinan Taia aeaa abai Where and How downloading and installing Basic one port device Linear Two Terminal Resistor o aaa es Linear Two Terminal Capacitor aasan eens Iwo Terminal Inductor 3 acech eek ei ese veo eee oa eK eae Ole wales Independent Voltage SOUICE aoaaa aaa Independent Current SOUrCE oe es Basic two port device Voltage Controlled Current Source aaau aaa Linear Voltage Controlled Voltage Source 1 ee Linear Current Controlled Current Source ow es Current Controlled Voltage SOUICE 26 es Linear Two Winding Ideal Transformer 0 0c a IASG NUMOF tese ice go he ist hk amt e nke eak eA ha omni i he wen de ih oh ae Coupled INQUCTIONS 64526008 stake ned hea ale eae ee a ae nae eae ae eae Polynomial TranscOnductor 0 ee ee eee eens semiconductor device JUNCTION DIOSAS saruri arer we Gah on Fok A pears wai KA war
89. ature coefficient of cj tpbsw Temperature coefficient of pbsw tcjsw Temperature coefficient of cjsw tpbswg Temperature coefficient of pbswg tcjswg Temperature coefficient of cjswg acde Exponential coefficient for finite charge thickness moin Coefficient for gate bias dependent surface potential noff C V turn on off parameter voffcv C V lateral shift parameter lint Length reduction parameter 11 Length reduction parameter llc Length reduction parameter for CV 11n Length reduction parameter 1w Length reduction parameter lwc Length reduction parameter for CV lwn Length reduction parameter 1lwl Length reduction parameter lwlc Length reduction parameter for CV Imin Real 0 0 Minimum length for the model used as a selection criterion when the model belongs to a cluster of models When the model does not belongs to a cluster and lt I gt is less than lt Imin gt an error is given The error can be a fatal one that cause the abortion of the simulation or a simply a warning one The severity of this error can be chosen through the lt bounderror gt parameter 1max Real 1 0 maximum length for the model used as a selection criterion when the model belongs to a cluster of models When the model does not belongs to a cluster and lt I gt 70 Chapter 4 Reference is greather than lt lmax gt an error is given The error can be a fatal one that cause the abortion of the simulation or a simply a warning one The severity of this
90. ault 1 003 Damping factor for temperature param 7 default 2 0e 6 Minimum temperature at which the method stops x simplex func x iters the simplex optimisation method tries to find the minimum of the func function This function must return a non negative real scalar The x vector stores the initial guess iters is the maximum allowed number of iterations that can be performed x sin y computes the sine of y The y argument can be a scalar a vector or a matrix X sort y sort in ascending order the content of the y real or complex vector sprintf sqrt std x y strptime str fmt convert the string str to the time structure x under the control of the format string fmt If fmt fails to match y is 0 otherwise y is set to the position of last matched character plus 1 Always check for this unless you re absolutely sure the date string will be parsed correctly Supported formats are those from the same C library procedure of Linux x sum y adds all elements of the y vector or matrix columnwise result is the x scalar or row vector respectively svd computes the singular values and left right singular vectors of a matrix system x transpose y computes the transpose of y Note this is NOT the conj transpose the conj transpose operator is represented by the standard symbol x zeros n m create a matrix of n rows and m colu
91. ax dX 1 312 m at lt N12 gt Max RelT 12 21 m at lt N21 gt where II R Il means the residue that is the norm of the right hand side vector of the steady state problem that is solved by employing the Newton iterative method A possible means 4 8 Advanced time domain analyses 127 that the sensitivity matrix has been computed D 10 tells that 10 entries of the updating vector computed by the Newton algorithm do not satisfy the convergence criterion R 10 tells that 10 equations do not meet the convergence criterion F 5 00000000 MHz shows the currently computed working frequency of the circuit obviously it has meaning only if the circuit is autonomous oscillator A possible tells that the algorithm was unable to suitably determine the working period and an old estimation is thus used eventually lt 4 gt shows that the transient analysis has been continued up to 4 times the previously estimated period in order to estimate a new and better one When the shooting method damps the updating the S 1 00e 01 notice appears showing the damping factor equal to 0 1 in this example Max dX 1 312 m is the maximum voltage difference obtained by subtracting voltages at the beginning and at the end of the circuit working period An indication of the circuit node at which the maximum difference has been computed is also given Max RelT 12 21 m is the maximum relative error is computing the initial conditio
92. ber of state variables used in the macromodel trtime Real 1 0e 9 this option has meaning only if the lt digital gt option is set true Digital devices make only discrete variations of their output voltages These variations are checked independently from their magnitudes each time there is a variation between two subsequent integration time instants this variation is forced to happen in less than lt trtime gt value In other words output voltage rise and falling times are ensured to be less than lt trtime gt This parameter helps the simulator to converge since when there is a transition of a lt digital gt VCVS the output voltage in theory must switch instantaneously From the simulator point of view this meas to shrink the integration time step till the minimum value allowed by the finite precision of the computer arithmetic unit Obviously this is time consuming since the length of the integration time step governs the simulation speed and at each transition of a digital vcvs the integration time step is reduced to its allowed minimum To avoid this drawback if the lt trtime gt is specified it defines the minimum integration time step below which the transition can be considered adequate MODEL DEFINITION model ModelName veys lt parameter value gt MODEL PARAMETERS macro Use device macromodel module Module name containing the macromodel evaluate The name of the macromodel evaluate routine args The name of the routine t
93. c Coefficient for SCC kvthOwe Threshold shift factor for well proximity effect k2we K2 shift factor for well proximity effect kuOwe Mobility degradation factor for well proximity effect scref Reference distance to calculate SCA wpemod Flag for WPE model WPEMOD 1 to activate this model lkvthOwe Length dependence of kvthOwe 1k2we Length dependence of k2we 1lkuOwe Length dependence of kuQwe wkvthOwe Width dependence of kvthOwe wk2we Width dependence of k2we wkuOwe Width dependence of ku0we pkvthOwe Cross term dependence of kvthOwe pk2we Cross term dependence of k2we pkuOwe Cross term dependence of kuQwe noia Flicker noise parameter noib Flicker noise parameter noic Flicker noise parameter tnoia Thermal noise parameter tnoib Thermal noise parameter rnoia Thermal noise coefficient rnoib Thermal noise coefficient ntnoi Thermal noise parameter em Flicker noise parameter ef Flicker noise frequency exponent af Flicker noise exponent kf Flicker noise coefficient nmos Flag to indicate NMOS 4 4 9 4 4 semiconductor device 93 pmos Flag to indicate PMOS cluster List list of bsim4 models composing the cluster A cluster does not describe a bsim4 model itself but groups of already defined bsim4 models Each model in the list is characterised by the lt wmin gt lt wmax gt lt Imin gt and lt lmax gt parameters When an instance refers to a cluster of models the list is searched and the current width and length
94. cit Cross term dependence of cit pnfactor Cross term dependence of nfactor pxj Cross term dependence of xj pvsat Cross term dependence of vsat pat Cross term dependence of at pad Cross term dependence of a0 pags Cross term dependence of ags pai Cross term dependence of al pa2 Cross term dependence of a2 pketa Cross term dependence of keta pnsub Cross term dependence of nsub pndep Cross term dependence of ndep pnsd Cross term dependence of nsd pphin Cross term dependence of phin pngate Cross term dependence of ngate pgammai Cross term dependence of gammal pgamma2 Cross term dependence of gamma2 pvbx Cross term dependence of vbx pvbm Cross term dependence of vbm pxt Cross term dependence of xt pki Cross term dependence of k1 pkt1 Cross term dependence of kt1 pkt11 Cross term dependence of kt11 pkt2 Cross term dependence of kt2 pk2 Cross term dependence of k2 pk3 Cross term dependence of k3 pk3b Cross term dependence of k3b pwO Cross term dependence of w0 pdvtpO Cross term dependence of dvtp0 pdvtpi Cross term dependence of dvtp1 plpeO Cross term dependence of IpeO 90 Chapter 4 Reference plpeb Cross term dependence of Ipeb pdvtO Cross term dependence of dvt0 pdvt1 Cross term dependence of dvt1 pdvt2 Cross term dependence of dvt2 pdvtOw Cross term dependence of dvtOw pdvtiw Cross term dependence of dvtlw pdvt2w Cross term dependence of dvt2w pdrout Cross term dependence of drout pdsub Cross term dependence of dsub
95. citor voltage in the first 20s left and from 700s to 720s right for the simple LC oscillator Netlist 2 3 2 panel but even in the first few periods show some differences After about 100 periods see Fig 2 9 right panel differences are dramatic The LC oscillator is conservative meaning that energy stored in the LC tank should remain constant in time Nevertheless the second order Gear algorithm slowly decays to zero The sixth order Gear algorithm does not decay to zero 2 4 2 4 1 2 4 Envelope time domain analysis envelope 27 but displays a relevant phase shift with respect to the Trapezoid simulation This phase shift is actually a different evaluation of frequency details on this frequency warping effect that is well known to astrophysicist and its impact in circuit simulation can be found in BS03 The frequency warping effect can be evidenced by adding to all simulation lines above the option findperiod yes This activates a rather simple but effective period computation algorithm Estimated oscillator periods found using the three different integration algorithms are significantly different as evidenced in Fig 2 10 PAN s time domain analysis has a large 6 35 6 34 F 6 33f r ATN PNE 6 321 a a NEEE K 6 31f 3 6 3 CL ee ee eee 6 28 6 277 Trap eas Gear 2 ial Gear 6 6 25 1 A 0 5 10 samples Figure 2 10 Oscillator period as evaluated by three different int
96. crgi First fitting parameter the bias dependent Rg xrcrg2 Second fitting parameter the bias dependent Rg lambda Velocity overshoot parameter vtl thermal velocity 1c back scattering parameter xn back scattering parameter vfbsdoff S D flatband voltage offset tvfbsdoff Temperature parameter for vfbsdoff tvoff Temperature parameter for voff lintnoi lint offset for noise calculation lint Length reduction parameter 11 Length reduction parameter llc Length reduction parameter for CV 11n Length reduction parameter 1w Length reduction parameter lwc Length reduction parameter for CV lwn Length reduction parameter 1lwl Length reduction parameter lwlc Length reduction parameter for CV Imin Minimum length for the model 1lmax Maximum length for the model wr Width dependence of rds wint Width reduction parameter dwg Width reduction parameter dwb Width reduction parameter wl Width reduction parameter wlc Width reduction parameter for CV wln Width reduction parameter ww Width reduction parameter wwe Width reduction parameter for CV wwn Width reduction parameter wwl Width reduction parameter wwlc Width reduction parameter for CV wmin Minimum width for the model wmax Maximum width for the model bO Abulk narrow width parameter bi Abulk narrow width parameter cgsl New C V model parameter cgd1 New C V model parameter ckappas S G overlap C V parameter ckappad D G overlap C V parameter cf Fringe capacitance parameter cle
97. cromodel in this case derivative values are specified in the macromodel The numerical method employed to implement the differentiator is the Gear one of order 1 and 2 delay Time delay stvar Number of state variable used in the macromodel MODEL DEFINITION model ModelName ceys lt parameter value gt MODEL PARAMETERS macro Use device macromodel module Shared module containing the macromodel evaluate The name of the macromodel evaluating routine Linear Two Winding Ideal Transformer This element implements an ideal transformer which implies that it works also at DC Winding 1 connects terminals w1 and w1 and similarly winding 2 connects w2 and w2 The number of turns on windings 1 and 2 are given respectively by t1 and t2 they must not be zero The absolute number of turns of each winding is not important only the ratio Both t1 and t2 are modifiable 1 is the default Current through winding 1 is computed Note that port 1 is described by using the normal convention for its voltage and current while port 2 is described with the non normal convention so that the v2 i2 product corresponds to the power flowing out from the port 2 i e v1 il v2 i2 0 To model a physical transformer with L1 and L2 being the inductance of the windings and K being the coupling coefficient add an inductor Lm K L1 in parallel with winding 1 and inductors Lel L1 1 K and Le2 L2 1 K in series with windings 1 and 2 respectively The turns ratio
98. d as O v Cxv cO 1 2 cl xv 1 3 c2 v where ck is the k th entry of the list Note that the default value of cO is 1 w Real capacitor width 1 Real capacitor length narrow Real 1 0 width correcting factor short Real 1 0 length correcting factor cj Real bottom capacitance per area cjsw Real sidewall capacitance per meter di Real 3 9 relative dielectric constant thick Real dielectric thickness tci Real 0 0 first order temperature coefficient tc2 Real 0 0 second order temperature coefficient tnom Real temperature at which the capacitor parameters have been defined Two Terminal Inductor The inductor is tied between the n1 and n2 terminals Inductor can be linear or nonlinear there are three different ways to describe a nonlinear inductor Note that a nonlinear inductor can be described only through a model card specification In the first case the nonlinear inductor is saturable and current controlled It is modeled by the relation Phi SNBmaxtanh NI LHmax SNmu0I L where Phi is flux S cross section of the core N number of turns L the average length of the core Hmax value of magnetic field at saturation Bmax value of the induction magnetic field at Hmax mu0 the vacuum permittivity and finally I inductor branch current As can be seen from the above formula nonlinearity is modeled by the tanh function When N I L gt Hmax the flux can be c
99. d by the frequencies given in the data file to find the frequencies used by the simulator Interpolation and extrapolation of the data is done using cubic splines on the data in polar form A simple minded algorithm is used to remove 2 pi jumps in the phase data This requires that the frequency points at which the data is measured must be close enough to avoid an excessive number of jumps Unfortunately noisy phase data can confuse the algorithm and result in unnecessary warning messages The file name is a quoted string It is very unwise to rely on extrapolated data The N port model supports also the definition of its electrical characteristic through the adoption of an external macromodel The external word means that it is a shared object run time loaded by the simulator and then suitably executed The electrical characteristic of the N port is descibed by the implicit equation G v Ai i tR 0 where G and Ai are a N x N matrices being N the port number v is the vector of port branch voltages i the vector of port branch currents and R is the N vector of known terms The routine describing the macromodel is called with the following ordered sequence of arguments lt input gt int an index that uniquely identifies the instance of the N port This index starts from 0 and it is incremented by 1 each time a new instance is found in the netlist lt input gt int number of ports of this N port lt input gt double the array v of the port
100. d gate drain junctions A good estimate for these capacitors is C pi epsilon w 2 where w is the gate width in microns and epsilon 0 116 fF micron ELEMENT DEFINITION Name drain gate source ModelName lt parameter value gt ELEMENT PARAMETERS area Junction area in square meters region DC operating region O off 1 on 2 rev 3 ohmic MODEL DEFINITION model ModelName gaas lt parameter value gt MODEL PARAMETERS nfet N type GaAs MESFET nmf N type GaAs MESFET pfet P type GaAs MESFET pmf P type GaAs MESFET vto Pinch off voltage beta Transconductance parameter lambda Channel length modulation parameter rd Real 0 01 drain resistance rs Real 0 01 source resistance is Gate saturation current js Gate saturation current 4 4 10 94 Chapter 4 Reference cgs Real 0 0 zero bias gate source junction capacitance cgd real 0 0 zero bias gate drain junction capacitance pb Gate junction potential fc Forward bias depletion capacitance threshold tnom Parameter measurement temperature kf Flicker noise coefficient af Flicker noise exponent imax Explosion current b Doping tail extending parameter alpha Saturation voltage parameter delta Gate capacitance pinch off transition width Philips juncap diode model The jucap diode model develope by Philips ELEMENT DEFINITION Name anode cathode ModelName lt parameter value gt ELEMENT PARAMETERS ab Real 1 0e 12 diffusion area area Real 1 0e 12 d
101. d in parallel nf Number of fingers sa distance between OD edge to poly of one side sb distance between OD edge to poly of the other side sd distance between neighbour fingers sca Integral of the first distribution function for scattered well dopant scb Integral of the second distribution function for scattered well dopant scc Integral of the third distribution function for scattered well dopant sc Distance to a single well edge min Minimize either D or S ad Drain area as Source area pd Drain perimeter ps Source perimeter nrd Number of squares in drain nrs Number of squares in source rbdb Body resistance rbsb Body resistance rbpb Body resistance rbps Body resistance rbpd Body resistance delvto Zero bias threshold voltage variation xgw Distance from gate contact center to device edge ngcon Number of gate contacts trngsmod Transient NQS model selector acngsmod AC NQS model selector rbodymod Distributed body R model selector rgatemod Gate resistance model selector geomod Geometry dependent parasitics model selector rgeomod S D resistance and contact model selector temp Device junction temperature noisetype Integer 7 types of generated noise possible choices are 0x0 noiseless BSIM4 0x1 thermal noise 0x2 shot noise 0x4 flicker noise The value of this parameter can be the result of the bitwise or of types of noise sources MODEL DEFINITION model ModelName bsim4 lt parameter value gt MODEL
102. d voltage BIN dd Vds field constant BIN dg Vds field constant BIN blk Leakage barrier lowering constant BIN clk Leakage scaling constant BIN iO Alias of clk BIN 1lkink Kink effect constant BIN mkink Kink effect exponent BIN mk Alias of mkink BIN vkink Kink effect voltage BIN rs Source resistance rd Drain resistance rsx Resistance in series with Cgs rdx Resistance in series with Cgd at DIBL parameter 1 BIN bt DIBL parameter 2 BIN eb Barrier height of diode BIN i00 Reverse diode saturation current BIN ioo Alias to 100 BIN etacO Capacitance subthreshold ideality factor at zero drain bias BIN etac00 Capacitance subthreshold coefficient of drain bias BIN mc Capacitance knee shape parameter BIN dvto Temperature coefficient of VTO BIN dmu1 Temperature coefficient of MU1 BIN dasat Temperature coefficient of ASAT BIN lasat Coefficient of length dependence of ASAT BIN von On Voltage BIN cgso Gate source overlap capacitance cgdo Gate drain overlap capacitance vsigma Above threshold DIBL parameter BIN vsigmat Above threshold DIBL parameter BIN me Long channel saturation transition parameter BIN ms Alias to me BIN minme Minimum value of me parameter meta ETA floating body parameter BIN 1s Channel length modulation coefficient 1 BIN vp Channel length modulation coefficient 2 BIN isubmod Channel length modulation model version
103. d wad a aR wees Bipolar Junction Transistor 2 ee eee ns Junction Field Effect Transistor n a auaa aaa Philios M903 mosfet transistor eens MOSFET level 2 device uuaa ee eee eens MOSFET level 3 device uuaa ne eee teens Berkeley Short Channel IGFET Model Version 3 2 0 0 0 0 ees Berkeley Short Channel IGFET Model Version 4 50 aaau aaaea aaa GGAS MESFET 02 c edu be edad be aA wt edn aed dd Roun ENEE EE Erkia Philips juncap diode Model es misc device analog to digital converter ees analog to digital converter kes CUITENT DIODE cristu ext anda WUE D acd atih EEEE adi EA aca a Microstrip LING oauan eee eee teens LINGO NI POM ii 2 5 cotton aad wate ed ne ah Wea tte whe hae hatch he Independent Resistive Source 1 ees Poly Silicon Thin Film Transistor 0 0 ce ee eens Time controled SWITCH weak a8 ee doe Ra Ew doe ee ee ee he de as TAMON LING cranto a cea wae hea oN wo dee eNO RA ee ole OS Voltage controlled switch 0 ees Basic analyses DG ANGIVSIS cvav eats ieee gana pam Veale G estel ale AG esa Da a snag be Small Signal ANAlysis es Time domain analysis aaa ee eee eas Simulator control pseudo analyses Set PON OPTIONS i iste ke TE ie Reb Ee a Oe ee ea bod ae aes Alter an instance parameter or simulation option 0 0 0 ee SWEEP ANGIYSIS oorti cca cali ahs Hades WEE Ea wack hain wack we COnNOLDIOCK eer wes there 6 a eet ek ee ak ae a ho beeen ne a 4 7 5 Monte Carlo Analysis aaan aaa aaa 122 4
104. damentals of the spectral analyses It is composed of two items The fist is the index the the harmonic of the first fundamantal and the second is the index of the harmonic of the second fundamental For example freq2t 1 1 defines the frequency 1 f1 1 f2 phaseit Real 0 0 phase of first fundamental in spectral analysis phase2t Real 0 0 phase of second time in the spectral analysis freq Real 0 0 frequency of the sinusoidal voltage source in time domain analyses such as tran shooring envelope and mtshooting vsin Real 0 0 maximum voltage of the sinusoidal source in time domain analyses such as tran shooring envelope and mtshooting ampl Real 0 0 maximum voltage of the sinusoidal source in time domain analyses such as tran shooring envelope and mtshooting vphase Degree 0 0 phase of the sinusoidal source in time domain analyses such as tran shooring envelope and mtshooting tdsin Real 0 0 the time interval along which the sinusoidal source is kept off The first period of the source beging exactly after this time interval more rigorously the generated waveform is v t vsin sin omega t tdsin vphase exp damping t tdsin damping Real 0 0 the damping factor of the sinusoidal source v t vsin sin omega t tdsin vphase exp damping t tdsin v Real 0 0 voltage value of a point in the pwl time sequence It is employed only in time domain analyses t Real 0 0 time instant of
105. de of these sources is swept stat stop and step values are automatically determined The user can modify these values by acting as in sweep analysis see the mealing of the related parameters 4 6 2 4 6 3 4 6 Basic analyses 109 Small Signal Analysis This analysis linearizes the circuit about the DC operating point and performs a small signal analysis If desired the analysis will be repeated while sweeping some independent variable The variable may be either frequency or some device instance parameter Not all device instance parameters are modifiable and some of those will change the DC operating point when modified If changing a parameter affects the DC operating point the operating point is recomputed at each step The modified parameter is reset to its original value after the analysis This analysis can determine also the poles and zeros of the linearised circuit In particular to compute zeros a transfer function must be defined by specifying the input source and the output voltage or current ANALYSIS DEFINITION Name ac lt parameter value gt ANALYSIS PARAMETERS annotate Integer default is global annotate level degree of annotation Values from 1 to 7 turn on display of generic information about the analysis run Values above 7 are bitwise and turn on specific printouts 0x08 print MNA matrix 0x10 dump MNA matrix in the analysis_name m file freq Real the frequency value when a device value parameter
106. dependence of cdscb wcdscd Width dependence of cdscd weit Width dependence of cit wnfactor Width dependence of nfactor wxj Width dependence of xj wvsat Width dependence of vsat wat Width dependence of at wa0 Width dependence of a0 wags Width dependence of ags wai Width dependence of al wa2 Width dependence of a2 wketa Width dependence of keta wnsub Width dependence of nsub wndep Width dependence of ndep wnsd Width dependence of nsd wphin Width dependence of phin wngate Width dependence of ngate 4 4 semiconductor device 87 wgammail Width dependence of gammal wgamma2 Width dependence of gamma2 wvbx Width dependence of vbx wvbm Width dependence of vbm wxt Width dependence of xt wk1 Width dependence of k1 wkt1 Width dependence of kt1 wkt11 Width dependence of kt11 wkt2 Width dependence of kt2 wk2 Width dependence of k2 wk3 Width dependence of k3 wk3b Width dependence of k3b wwO Width dependence of wO wdvtpO Width dependence of dvtp0 wdvtp1 Width dependence of dvtp1 wlpeO Width dependence of lpe0 wlpeb Width dependence of lpeb wdvt0O Width dependence of dvtO wdvt1 Width dependence of dvt1 wdvt2 Width dependence of dvt2 wdvtOw Width dependence of dvt0w wdvtiw Width dependence of dvtlw wdvt2w Width dependence of dvt2w wdrout Width dependence of drout wdsub Width dependence of dsub wvthO Width dependence of vto wvthO Width dependence of vto wua Width dependence of ua wua1 Width dependence of ual wub Width de
107. ductor corelength Magnetic core length meter It is the average length of core and is meaningful only for a nonlinear saturable inductor hmax Maximum magnetic field amper meter When the magnetic field is above this value the inductor is considered saturated It is meaningful only for a saturable inductor turns Number of turns It is meaningful only for a nonlinear saturable inductor cO Coefficient of the modeling polynomial of nonlinear inductor c1 Coefficient of the modeling polynomial of nonlinear inductor c2 Coefficient of the modeling polynomial of nonlinear inductor c3 Coefficient of the modeling polynomial of nonlinear inductor c4 Coefficient of the modeling polynomial of nonlinear inductor c5 Coefficient of the modeling polynomial of nonlinear inductor Independent Voltage Source The positive node is n and the negative node is n Current through the source is computed and is defined to be positive if it flows from the positive node through the source to the negative node The only meaningful parameter in a DC analysis is vdc In an AC analysis only mag and phase are meaningful In a harmonic analysis vdc mag1 mag2 phasel phase2 etc are used In a spectral analysis vdc mag1f phaself mag2f and phase2f are used In the time domain analyses such as transient shooting envelope multishooting the piece wise linear source pulse 46 Chapter 4 Reference source sinusoidal and exponential ones are meaningfull In a
108. e capacitors and inductors that specify state variables These state variables are monitored by the procedure that determines the working period of the autonomous circuit If this list is not given all state variables are monitored save String the path of the file into which the solution is saved bz2 format is adopted The solution stored in this file can be later loaded by the simulator and used as initial condition from which the envelope following analysis is started load String the path of the file from which a previously saved solution is read The file format can be both ascii and bz2 and it is automatically determined during loading This 124 Chapter 4 Reference solution is used as initial condition to the envelope following analysis acntrl Integer 1 select the accuracy control method 1 relative energy balance this method computes the electrical energy variation of each capacitor and inductor along each integration interval and the corresponding electrical work delivered by the resistive one port elements to which they are connected The electrical work and energy are compared and the integration time step is chosen in order to keep their difference below the user defined threshold 2 absolute energy balance This method is identical to the previous one but the difference between the electrical work and energy is compared to the total energy stored into the corresponding capacitor inductor The integration time step is
109. e do not discuss here some of the messages displayed in the logfile 2 3 2 3 1 2 3 The time domain analysis tran 23 since they can be ignored unless convergence problems arise The time domain analysis tran The tran analysis computes the evolution in time of a circuit Since it is based on MNA it will actually compute the evolution of all node voltages and only of currents in bad branches A Van der Pol oscillator parameters expressions and the poly device Consider the Van der Pol oscillator shown in Fig 2 6 whose netlist is shown in 2 3 1 Figure 2 6 A Van der Pol oscillator Netlist 2 3 1 Multiple DC solution File vanderpol pan ground electrical 0 parameters R 10 ALPHA 1 3 Analyses TranO tran tstop 100 uic 1 Trani tran tstop 120 restart no 3 Circuit ri n20 O n20 0 poly n0 0 ni 1 R n2 0 n3 ALPHA R d0 1 vri ni0 n20 vsource vdc 0 c1 n10 O capacitor c 1 icon 3 11 n10 O inductor l 1 Before we perform the simulation and discuss the tran analysis line let s briefly comment some interesting features of this netlist Parameters and Expressions First of all note that all numerical values can be defined using parameters parameters R 10 ALPHA 1 3 It is possible to define as many parameters as desired parameters can be defined as a mathemati cal expression that may contain other parameters Any parameter can then be used in the netlist and combined in exp
110. e one node at each port grounded There are several different ways to specify the length of a line First is to simply give the length and velocity recall that velocity 1 sqrt relative dielectric constant The second is to give the reference frequency and the normalized length The last is to give the delay in seconds There are three mutually exclusive ways to specify dielectric shunt loss g alphad and qd Both alphad and qd are low loss approximations Similarly there are three ways to specify conductor series loss r alphac and qc All imply a skin effect loss that is proportional to the square root of frequency and so require the use of fc If dcr is given then the conductor is assumed to be of finite thickness and so the loss does not go to zero at DC but rather approaches some constant value as frequency decreases below the corner frequency It is also possible to specify loss by giving the DC resistance and the corner frequency It is assumed that the loss from skin effect was measured at a frequency well above the corner frequency This problem can be avoided by specifying loss with dcr and corner ELEMENT DEFINITION Name port portl port2 port2 ModelName lt parameter value gt ELEMENT PARAMETERS z0 Characteristic impedance td Time delay f Reference frequency nl Normalized length in wavelengths at f vel Propagation velocity normalized to c len Physical length in meters 4 5 10 104 Chapter 4 Reference
111. each time interval are also computed by adopting the dalay operator in the frequency domain It relates each solution at the end of a time interval to those at the beginning of all the time intervals eabstol Absolute tolerance employed by the multi tone shooting analysis default 1 0e 12 ereltol Relative tolerance employed by the multi tone shooting analysis default 1 0e 3 iabstol Absolute current tolerance employed by the Newton algorithm at each time point of the time domain analysis default 1 0e 12 ireltol Relative current tolerance employed by the Newton algorithm at each time point of the time domain analysis maxiter Maximum number of allowed iterations for the multi shooting algorithm default 100 maxdeltai Real maximum allowed variation of current of state variables at each iteration of the Newton algorithm If variation of any of the state variabled exceeds the maximum allowed value the Newton step is suitably shrinked preserving direction maxdeltav Real maximum allowed variation of voltage of state variables at each iteration of the Newton algorithm If variation of any of the state variabled exceeds the maximum allowed value the Newton step is suitably shrinked preserving direction method Select the integration method 1 trapezoidal default 2 Gear method of various order order Integer 2 set the order of the Gear integration method It has no meaning if other integration methods are employed harms Integ
112. eal the unique voltage threshold used to define the switching threshold of a 1 bit converter trtime Real the minimum allowed time window into which a variation of the digital output generated by the a2d conveter must fall During a time domain analysis the integration time step is shortened such that any bit change of the coding of the digital net corresponding to a variation of the analog voltage falls inside a time window of length less than the value of this parameter Current probe Current probe is a pseudo element employed exclusively by the harmonic and the spectral analyses It behaves as an independet current source at the specified harmonics and as a short circuit at the remaining harmonics of the spectrum ELEMENT DEFINITION Name n n iprobe lt parameter value gt ELEMENT PARAMETERS pimagi Real 0 0 the magnutude of the current probe In case of the lt spectral gt analysis this probe acts on the first fundamental and its harmonics pipht Real 0 0 the phase of the current probe In case of the lt spectral gt analysis this probe acts on the first fundamental and its harmonics pigi Real 0 0 the value of the conductance connected in parallel to the current probe In case of the lt spectral gt analysis this probe acts on the first fundamental and its harmonics freq2t List the list of indices that defines the frequency of the second probe as a linear combination of the two fundamental
113. ectrical variable followed by the spectrum postfix Therefore spectra can be retrieved through the command get TranName node spectrum 3 When the parameter to compute the working period of a periodic circuit is turned on these other results are available period the computed working period of the circuit tstart the starting time instance of a computed period tstop the last time instant of the computed period 110 Chapter 4 Reference samples the index identifying the working period x axis of these results 4 The monodromy matrix of the circuit can be retrieved through the command get TranName monodromy The vector that maps the column index of the nonodromy matrix to the corresponsing MNA equation can be retrived through the command get TranName eqmap 5 Results by time domain noise analysis can be retrived through the command get TranName node noise Results are computed on a regular time point grid Time instants can be retrieved through the command get TranName noisetime ANALYSIS DEFINITION Name tran lt parameter value gt ANALYSIS PARAMETERS annotate Degree of annotation It is an integer number values from 1 to 7 turn on displaying of generic informations about the analysis run Values above 7 are bitwise and turn on specific printouts In particular 0x08 print some information about the convergence of the Newton method 0x10 print parti
114. ed In the body a dc analysis is first executed the count variable is increased by 1 and then the voltage al node x is retrieved and stored in an entry os the samples array When the MonteCarlo analysis is completed the mean value and the standard deviation of the sampled stored in the array are computed and displayed ANALYSIS DEFINITION Name me lt parameter value gt 4 8 4 8 1 4 8 Advanced time domain analyses 123 ANALYSIS PARAMETERS annotate Integer 3 degree of annotation clip Boolean 1 allow clipping of instance and model parameters when their values generated by the ramdom processe go above the maximum allowed value and below the minimum allowed one samples Integer 100 number of samples considered in the MonteCarlo analysis seed Integer the seed of the random number generators If not specified the seed is taken from the system clock maxsigma Integer 4 define the maximum variance below which a randomly generated number must fall If the random number is above this variance it is dropped Advanced time domain analyses Time domain analysis with envelope algorithm The envelope analysis computes the time domain behavior of both autonomous and non autonomous circuits Waveforms made available to the control analysis are 1 Node voltages and branch currents computed by the envelope analysis along each time window time samples are stores in the time vector These waveform
115. ed samples fi Boolean false turn on and off the computation of the Fourier integrals i e the computation of the components of the waveform spectra fifund Real the frequency of the fundamental A fraction of the reciprocal of this value time period is employed to sample waveforms For more details see the lt fiperiod gt parameter fiperiod The period along which Fourier integrals of the computed wavefors are performed The interval should begin lt fiperiod gt before the lt tstop gt value If the lt fiestper gt parameter is set true the lt fiperiod gt value is automatically adjusted within a maximum value of 40 fiestper Boolean false estimate the period of waveforms to which Fourrier integrals are applied If this parameter is true it is not mandatory to specify an accurate value of lt ifperiod gt since it is automatically adjusted within a maximum relative value of 40 fifreq List list of frequecies at which the Fourier integrals are computed findperiod Boolean false find period values during the tran analysis This can be used to monitor synchronization transients Unconsistent results are possibly found if a nonperi odic dynamics is present Period doubling may also cause problems monodromy Boolean false compute the monodromy matrix of the circuit printmo Boolean false print the monodromy matrix of the circuit jump Boolean false activate finding of discontinuities in the vector field T
116. efficient 1 kt1 Temperature coefficient of Vth kt11 Temperature coefficient of Vth kt2 Body coefficient of kt1 k2 Bulk effect coefficient 2 k3 Narrow width effect coefficient k3b Body effect coefficient of k3 wO Narrow width effect parameter nlx Lateral non uniform doping effect dvtO Short channel effect coeff 0 dvti Short channel effect coeff 1 dvt2 Short channel effect coeff 2 dvtOw Narrow Width coeff 0 dvtiw Narrow Width effect coeff 1 dvt2w Narrow Width effect coeff 2 drout DIBL coefficient of output resistance dsub DIBL coefficient in the subthreshold region vthO Threshold voltage vtho Threshold voltage ua Linear gate dependence of mobility ual Temperature coefficient of ua ub Quadratic gate dependence of mobility ubi Temperature coefficient of ub uc Body bias dependence of mobility uci Temperature coefficient of uc u0 Low field mobility at Tnom ute Temperature coefficient of mobility voff Threshold voltage offset tnom Parameter measurement temperature cgso Gate source overlap capacitance per width cgdo Gate drain overlap capacitance per width cgbo Gate bulk overlap capacitance per length xpart Channel charge partitioning elm Non quasi static Elmore Constant Parameter delta Effective Vds parameter rsh Source drain sheet resistance rdsw Source drain resistance per width prwg Gate bias effect on parasitic resistance prwb Body effect on parasitic resistance prt Temperature coefficient of parasitic r
117. efficient for pocket devices 80 Chapter 4 Reference pdits Coefficient for drain induced Vth shifts pditsl Length dependence of drain induced Vth shifts pditsd Vds dependence of drain induced Vth shifts pscbe1 Substrate current body effect coefficient pscbe2 Substrate current body effect coefficient pvag Gate dependence of output resistance parameter jss Bottom source junction reverse saturation current density jsws Isolation edge sidewall source junction reverse saturation current density jswgs Gate edge source junction reverse saturation current density pbs Source junction built in potential njs Source junction emission coefficient xtis Source junction current temperature exponent mjs Source bottom junction capacitance grading coefficient pbsws Source sidewall junction capacitance built in potential mjsws Source sidewall junction capacitance grading coefficient pbswgs Source gate side sidewall junction capacitance built in potential mjswgs Source gate side sidewall junction capacitance grading coefficient cjs Source bottom junction capacitance per unit area cjsws Source sidewall junction capacitance per unit periphery cjswgs Source gate side sidewall junction capacitance per unit width jsd Bottom drain junction reverse saturation current density jswd Isolation edge sidewall drain junction reverse saturation current density jswgd Gate edge drain junction reverse saturation current density pbd Drain junction built in potent
118. egration algorithms number of other options more than 70 as always a list and brief description of these options can be found with command pan s tran Envelope time domain analysis envelope The Envelope time domain analysis performs a fast time domain analysis using information given by the envelope of the solution waveforms Waveforms resulting from this analysis appear to jump in time rapidly converging towards a stable steady state solution Results of the envelope analysis can be used to restart other time domain simulations and is useful to quickly arrive in proximity of a steady state solution The Bubba oscillator using Schematic Capture and Netlisting tools Consider now a simple sine wave oscillator specifically a modified and simplified version of the Bubba oscillator found at page 17 of MP01 Since we do not have a netlist for this oscillator we will draw one using the standard open source tool gschem that is part of the geda tool suite The schematic of the Bubba oscillator is shown in Fig 2 11 note that a SPICE include has been added so that the model of the TLV247xA op amp is included in the final spice netlist This schematic can be transformed in a spice netlist using gnetlist another tool in the geda suite This tool includes a number of different backends that can be used to process a schematic in many different ways The spice backend that we will use is spice sdb the older backend spice is discont
119. elements to which they are connected The electrical work and energy are compared and the integration time step is chosen in order to keep their difference below the user defined threshold 2 absolute energy balance This method is identical to the previous one but the difference between the electrical work and energy is compared to the total energy stored into the corresponding capacitor inductor The integration time step is chosen to keep variation with respect to total stored energy below the user defined threshold In general this method gives less accurate results with respect to the previous one but the analysis runs faster ntoltype Integer 2 this option specifies the tolerance check adopted at each time point of the analysis to control if the Newton method has reached convergence Checks are done both on the variation of variables computed by the Newton method and on the entries of the right hand side vector Define as dX n the variation of the n electrical variable as reltol and abstol the relative and absolute maximum allowed variations respectively The n th variable has converged if dX n lt reltol Y n abstol The ntoltype parameter defines three different ways to compute Y n number as 1 2 and 3 Their meanings are listed in the following 1 global Y n is defined as the maximum over the time interval till the current time point of the maximum value assumed by the X n
120. em unknowns and uses the Newton iterative method to solver the enlarge harmonic balance problem This method can not be employed by itself since it needs an initial guess This guess can be obtained for example by a previous harmonic analysis that employs a probe but that does not solve it or by time domain analyses such as trasient envelope shooting freltol Real 1 0e 6 the relative error tolerance in the computation of the working frequency of autonomous circuits fabstol Real 1 0e 12 the absolute error tolerance in the computation of the working fre quency of autonomous circuits floquet Boolean false compute the Floquet multipliers of the circuit 4 9 3 140 Chapter 4 Reference flgharms Real the number of harmonics employed to determine the Floquet multipliers clsminsize Integer the minimum number of elements that have to populate a cluster in order to have the cluster itself considered as meaningful During clustering of Floquet exponents there can be clusters composed of a small number of elements with respect to the total number of harmonics This is caused by spectrum truncation and round off errors This parameter introduces a threshold so that these clusters are ignored vprobe List the list containing the pair of nodes at which the voltage probe is connected If the second node of the list is omitted it is considered connected to ground The format is lt vprobe pos neg gt vprobemagi
121. ence of beta0 wagidl Width dependence of agidl wbgidl Width dependence of bgidl wcgidl Width dependence of cgidl wegidl Width dependence of egidl waigc Width dependence of aigc wbigc Width dependence of bigc wcigc Width dependence of cigc waigsd Width dependence of aigsd wbigsd Width dependence of bigsd wcigsd Width dependence of cigsd waigbacc Width dependence of aigbacc wbigbacc Width dependence of bigbacc wcigbacc Width dependence of cigbacc waigbinv Width dependence of aigbinv wbigbinv Width dependence of bigbinv wcigbinv Width dependence of cigbinv wnigc Width dependence of nigc wnigbinv Width dependence of nigbinv wnigbacc Width dependence of nigbacc wntox Width dependence of ntox weigbinv Width dependence for eigbinv wpigcd Width dependence for pigcd wpoxedge Width dependence for poxedge 4 4 semiconductor device 89 wvfbcv Width dependence of vfbcv wvfb Width dependence of vfb wacde Width dependence of acde wmoin Width dependence of moin wnoff Width dependence of noff wvoffcv Width dependence of voffcv wxrcrgi Width dependence of xrcrg1 wxrcrg2 Width dependence of xrcrg2 wlambda Width dependence of lambda wvtl Width dependence of vtl wxn Width dependence of xn weu Width dependence of eu wvfbsdoff Width dependence of vfbsdoff wtvfbsdoff Width dependence of tvfbsdoff wtvoff Width dependence of tvoff pcedsc Cross term dependence of cdsc pcdscb Cross term dependence of cdscb pcedscd Cross term dependence of cdscd p
122. ency grid fs is the sampling frequency printf q 1 p qr a computes the q and r decomposition of the a matrix entries on the diagonal of r ordered in decreasing magnitude p is a permutation matrix such that q r 4 7 Simulator control pseudo analyses 121 a p 0 x rand rows cols fills x with pseudo random elements uniformly distributed between 0 and 1 x randn rows cols fills x with normally distributed pseudo random elements having zero mean and variance one real x round y return the integer nearest to y If y is complex return x round real y i round imag y shift x siman func x param the simulated annealing optimisation method tries to find the minimum of the func function This function has as argument the x vector of parameters versus which the minimum is looked for The func function must return a positive scale representing the value assumed by the cost functions The param argument is a vector whose entries are the 7 parameters governing the behaviour of the simulated annealing optimisation method The measing of each entry follows param 1 default 200 How many points do we try before stepping param 2 default 1000 How many iterations for each T param 3 default 1 0 maximun step size in random walk param 4 default 1 0 Boltzmann constant param 5 default 0 008 Initial temperature param 6 def
123. endence of cit pnfactor Cross term dependence of nfactor pxj Cross term dependence of xj pvsat Cross term dependence of vsat pat Cross term dependence of at pad Cross term dependence of a0 pags Cross term dependence of ags pai Cross term dependence of al pa2 Cross term dependence of a2 pketa Cross term dependence of keta pnsub Cross term dependence of nsub pnch Cross term dependence of nch pngate Cross term dependence of ngate pgammai Cross term dependence of gammal pgamma2 Cross term dependence of gamma2 pvbx Cross term dependence of vbx pvbm Cross term dependence of vbm pxt Cross term dependence of xt pki Cross term dependence of k1 pkt1 Cross term dependence of kt1 pkt11 Cross term dependence of kt11 pkt2 Cross term dependence of kt2 pk2 Cross term dependence of k2 pk3 Cross term dependence of k3 pk3b Cross term dependence of k3b pwO Cross term dependence of w0 pnix Cross term dependence of nlx pdvtO Cross term dependence of dvt0 pdvt1 Cross term dependence of dvt1 pdvt2 Cross term dependence of dvt2 pdvtOw Cross term dependence of dvtOw pdvtiw Cross term dependence of dvtlw pdvt2w Cross term dependence of dvt2w pdrout Cross term dependence of drout pdsub Cross term dependence of dsub pvtho Cross term dependence of vto 4 4 semiconductor device 75 pvthO Cross term dependence of vto pua Cross term dependence of ua puai Cross term dependence of ual pub Cross term dependence of ub pub1 Cross term dependence of
124. ent source or simulator options e g temperature may be swept Sweep among other applications can be used to plot the DC characteristic of a non linear component In this case we wish to plot the characteristic of an instance of a diode described by some of its process parameters We thus add the model line model mydiode diode is 3 142f rs 3 536 n 1 xti 3 eg 1 11 m 0 4186 vj 0 75 fc 0 5 bv 80 ibv 24 573m imax 100 where the model of a device of type diode and named mydiode is created A quick reference to 2 2 The DC analysis 17 the meaning of diode parameters in the line above is given by command pan s diode All these parameters are compatible with those found in standard commercial component libraries Now mydiode can be used as any other diode component More than one diode model can be specified in each netlist file Netlist 2 2 4 Simple netlist File Simple_002 pan ground electrical 0 Analyses DcO de print yes instance vi param vdc start 85 0 step 0 01 stop 10 0 Circuit vi ni 0 vsource vdc 2 di ni 0 mydiode Models model mydiode diode is 3 142f rs 3 536 n 1 xti 3 eg 1 11 m 0 4186 vj 0 75 fc 0 5 bv 80 ibv 24 573m imax 100 The DcO analysis line instructs the simulator to perform a large number of independent DC analysis sweeping parameter vdc of instance v1 from 85 0V to 10 0V with steps of 100mV The resulting output imported and plotted using Matlab
125. er 2 largest considered harmonic of the first and second frequency printnodes Print external internal node map restart If set to false results obtained by a previous envelope or transient analysis are uses as initial guess to the shooting method spectra Compute spectra of time domain waveforms time Boolean 0 output time domain waveforms by starting from the computed spectra points Integer 1000 number of time points in output waveforms thermalnet Use the thermal network tinc Real 1 1 lengthening factor of the time step at convergence tmax Maximum allowed time step lenght tmin Minimum allowed time step lenght 4 8 4 134 Chapter 4 Reference trabstol Absolute convergence tolerance of the transient analysis employed by the envelope analysis trreltol Relative convergence tolerance of the transient analysis employed by the envelope analysis savelist It defines a list on nodes or branch currents that will be saved into the output file during simulation Other computed electrical variables are not saved uic Integer 0 use given initial conditions of inductors and capacitors Initial conditions are considered in two different ways according to the value of this paramenter in particular 1 it is assumed that all the capacitors and inductors have suitable initial conditions This means that if they are not specified by the user they are assumed equal to zero Note that this initial condition determination does not dea
126. erature i e 27 Celsius the second at T 10 and the third at T 120 Results zoomed in to show the right knee area are shown in Fig 2 3 The alter statement in netlist 2 2 5 modifies the value of a global simulation option Alteri alter opt yes param temp value 10 device parameters are modified using a similar syntax to modify resistor instance r1 a possible alter card is shown below ri ni n2 resistor r 1k Alter_ri alter instance ri param r value 10k Recall that a list of options and parameters can be quickly obtained using command pan s name where name is the keyword options for global simulator options or the name of the device or of the analysis for specific options 2 2 The DC analysis 19 0 27 f 0 17 0 07 0 03 i i 0 Figure 2 3 Diode current vs generator voltage at three different temperatures generated by Netlist 2 2 5 2 2 4 Multiple DC sweeps In many cases one may desire to sweep two different circuit parameters at the same time this is possible by simply adding more than one sweep instruction in the same analysis any number of embedded sweeps are possible even if possibly impractical and difficult to analyse A simple circuit where two parameters are swept at the same time is shown in Netlist 2 2 6 Netlist 2 2 6 Double DC sweep File Simple_004 pan ground electrical 0 Analyses DcO de print yes instance v1 p
127. ercial simulators search for multiple operating points is not implemented Consider the voltage regulator shown in Fig 2 5 whose netlist is shown in 2 2 7 _Vad ji SRI R2 ae j 93 RBS Q G A es pal 9 3 a ad ad Figure 2 5 A simple voltage regulator 2 2 The DC analysis 21 Netlist 2 2 7 Multiple DC solution File vregu pan Voltage regulator ground electrical gnd Analyses Find possible multiple solutions Del dc print yes multiple yes stab 1 3 Circuit VCC VCC gnd vsource vdc 20 ri VCC a resistor r 5k qi b b a vec P r2 VCC l1 resistor r 5k q2 De VEG P q6 b d e gndN qT c d f gnd N area 5 r3 f e resistor r 1k r4 e gnd resistor r 5k q3 ae e Came q5 g d o gndN r5 o gnd resistor r 5k q4 vcc g out gnd N r6 out d resistor r 3 54k r7 d gnd resistor r 1k ifdef STARTUP Startup q8 c i e gndP q9 i i gnd gnd N r8 VCC i resistor r 100K endif Models model P bjt pnp yes bf 200 vaf 50 model N bjt npn yes bf 600 vaf 100 The circuit is a simple voltage regulator two simple BJT models are used the remaining part of the netlist is quite simple and straightforward except for the part enclosed in the two statements ifdef STARTUP devices analyses options endif PAN allows the use of Simulator directives These instructions allow conditional inclusion of device analysis and option declarations I
128. ergence tolerance of the transient analysis employed by the envelope analysis fft Boolean flag that if set true turns on FFT analysis of electrical variables default false Fundamental frequency corresponds to the circuit working period given by the user if the circuit is non autonomous and computed by the envelope algorithm if the circuit is autonomous fftharms Number of harmonics considered by FFT fftsamples The number of sampling points in the FFT analysis FFT analysis ensures that the given number of points is employed but more points may be employed This parameters is effective only if it speficies a number of samples grether than 4 lt ffthars gt 1 which represents the minimum allowed number of employed samples floquet It is a boolen parameter default false If set true the Floquet multipliers related to the jacobian of the state transition matrix computed at convergenge of the shooting method are evaluated lyapunov It is a boolen parameter default false If set true the Lyapunov exponents are 4 8 2 126 Chapter 4 Reference evaluated maxiter Integer 100 maximum number of allowed iterations for the envelope algorithm minper Set the minimum value assumed by the estimated circuit working period This parameter has meaning only if the lt autonomous gt parameter is set to true cmin A parasitic capacitor of value lt cmin gt is added between each node user given nodes of the circuit and ground Thi
129. eries It is ensured that at least the specified number of points are used but more points can be employed The coefficients are computed by evaluating the Fourier integrals and not by applying the FFT algorithm This ensures better results even though few coefficients are computed This parameters is effective only if it specifies a number of samples greater than 4 lt ffthars gt 1 which represents the minimum allowed number of employed samples method Integer 1 select the integration method 1 trapezoidal 2 Gear methods of order from 1 to 6 order Integer 2 set the order of the Gear integration method It has no meaning if other integration methods are employed minper Set the minimum value assumed by the estimated circuit working period This parameter has meaning only if the lt autonomous gt parameter is set to true uic Integer 0 use given initial conditions of inductors and capacitors Initial conditions are considered in two different ways according to the value of this parameter in particular 1 it is assumed that all the capacitors and inductors have suitable initial conditions This means that if they are not specified by the user they are assumed equal to zero Note that this initial condition determination does not deals with loops of voltages source and capacitors even though consistent initial conditions are given since the loop current is under determined The same applies to inductor cut sets 2 itis assumed th
130. erm dependence of noff pvoffcv Cross term dependence of voffcv pxrcrgi Cross term dependence of xrerg1 pxrcerg2 Cross term dependence of xrerg2 plambda Cross term dependence of lambda pvtl Cross term dependence of vtl pxn Cross term dependence of xn peu Cross term dependence of eu pvfbsdoff Cross term dependence of vfbsdoff ptvfbsdoff Cross term dependence of tvfbsdoff ptvoff Cross term dependence of tvoff saref Reference distance between OD edge to poly of one side 92 Chapter 4 Reference sbref Reference distance between OD edge to poly of the other side wlod Width parameter for stress effect ku0 Mobility degradation enhancement coefficient for LOD kvsat Saturation velocity degradation enhancement parameter for LOD kvthO Threshold degradation enhancement parameter for LOD tku0 Temperature coefficient of KUO llodkuO Length parameter for u0 LOD effect wlodku0 Width parameter for u0 LOD effect llodvth Length parameter for vth LOD effect wlodvth Width parameter for vth LOD effect 1ku0 Length dependence of ku0 wku0 Width dependence of ku0 pku0 Cross term dependence of ku0 1lkvthO Length dependence of kvth0O wkvthO Width dependence of kvthO pkvth0O Cross term dependence of kvth0O stk2 K2 shift factor related to stress effect on vth lodk2 K2 shift modification factor for stress effect steta0 eta0 shift factor related to stress effect on vth lodeta0 etaO shift modification factor for stress effect web Coefficient for SCB we
131. error can be chosen through the lt bounderror gt parameter wr Width dependence of rds wint Width reduction parameter dwg Width reduction parameter dwb Width reduction parameter wl Width reduction parameter wlc Width reduction parameter for CV wln Width reduction parameter ww Width reduction parameter wwe Width reduction parameter for CV wwn Width reduction parameter wwl Width reduction parameter wwlc Width reduction parameter for CV wmin Real 0 0 minimum width for the model used as a selection criterion when the model belongs to a cluster of models When the model does not belongs to a cluster and lt w gt is less than lt wmin gt an error is given The error can be a fatal one that cause the abortion of the simulation or a simply a warning one The severity of this error can be chosen through the lt bounderror gt parameter wmax Real 1 0 maximum width for the model used as a selection criterion when the model belongs to a cluster of models When the model does not belongs to a cluster and lt w gt is less than lt wmax gt an error is given The error can be a fatal one that cause the abortion of the simulation or a simply a warning one The severity of this error can be chosen through the lt bounderror gt parameter bO Abulk narrow width parameter bi Abulk narrow width parameter cgsl New C V model parameter cgd1 New C V model parameter ckappa New C V model parameter cf Fringe capacitance parameter clc Vdsa
132. esign EnvO envelope tstop 200m period 1 5m autonomous yes uic 1 Tran0 tran tstop 210m tmax 0 01m restart no Vss_s Vss 0 2 5 Vdd_s Vdd 0 5 0 C4 0 n5 10n C3 0 ni 10n C2 O sin_out 10n ic 2 R4 sin_out n7 10k R6 n5 cos_out 10k Ri n5 1 360k R2 1 n4 1 5Meg C1 O n3 10n R5 ni n2 10k R3 n4 n3 10k X4 sin_out n2 Vdd O n2 TLV247xA X2 n3 n7 Vdd O n7 TLV247xA X3 ni cos_out Vdd O cos_out TLV247xA Xi Vss 1 Vdd O n4 TLV247xA end Note that in our case a strict ordering is required since we wish to perform a quick envelope simulation followed by a short tran analysis that restarts from the previous envelope results note the restart no option of analysis Tran0 The Env0 simulation line i e EnvO envelope tstop 200m period 1 5m autonomous yes uic 1 instructs PAN to perform an envelope analysis starting at time 0 ending at 200ns with an approximate oscillator period of 1 5ms on an autonomous circuit i e without external periodic inputs different algorithms are used in the two cases and finally using the initial conditions given in the netlist Results of this analysis are shown in Fig 2 12 Envelope analysis only performs few cycles of the fast dynamics at a time predicting and simulating new cycles further in time only when needed This is shown in the left panel A brief transient analysis follows Details of the last envelope sinusoids followed by the initial transient ones are shown
133. esistance eta0 Subthreshold region DIBL coefficient etab Subthreshold region DIBL coefficient 4 4 semiconductor device 69 pclm Channel length modulation Coefficient pdiblci Drain induced barrier lowering coefficient pdiblc2 Drain induced barrier lowering coefficient pdiblcb Body effect on drain induced barrier lowering pscbe1 Substrate current body effect coefficient pscbe2 Substrate current body effect coefficient pvag Gate dependence of output resistance parameter js Source drain junction reverse saturation current density jsw Sidewall junction reverse saturation current density pb Source drain junction built in potential nj Source drain junction emission coefficient n Source drain junction emission coefficient xti Junction current temperature exponent mj Source drain bottom junction capacitance grading coefficient pbsw Source drain sidewall junction capacitance built in potential mjsw Source drain sidewall junction capacitance grading coefficient pbswg Source drain gate side sidewall junction capacitance built in potential mjswg Source drain gate side sidewall junction capacitance grading coefficient cj Source drain bottom junction capacitance per unit area vfbcv Flat Band Voltage parameter for capmod 0 only vfb Flat Band Voltage cjsw Source drain sidewall junction capacitance per unit periphery cjswg Source drain gate side sidewall junction capacitance per unit width tpb Temperature coefficient of pb tcj Temper
134. eter can be changed with each alter statement This statement can be employed also to change a modifiable global simulation option like temperture It effects any following analysis In general when a device parameter or a global simulation options is changed the results computed by previous analyses those that preceded the current alter stetement become invalid that is they can no be futher used to start the following analyses ANALYSIS DEFINITION Name alter lt parameter value gt ANALYSIS PARAMETERS instance String the name of the instance whose parameter is modified model String the name of the model whose parameter is modified opt Boolean specifies that a parameter of simulation options is modified param String the name of parameter to be altered value Real 0 0 the new value assigned to the parameter annotate Integer 3 degree of annotation invalidate Boolean true the results computed by the analyses carried out before the appli cation of the current lt alter gt statement are invalidated Therefore they can not be further used as a starting guess for the following analyses Sweep Analysis This analysis sweeps parameters of a model an instance or a simulator option and executes the actions specified in the action body Sweeps can be applied also to parameters defined through the parameter keyword in the circuit netlist More that one sweep can be defined in the same sweep card For ex
135. evaluation of the initial conditions opvabstol Absolute voltage tolerance for the evaluation of the initial conditions opvreltol Relative voltage tolerance for the evaluation of the initial conditions maxdeltav Real the maximum allowed variation of node voltage allowed during the iterations of the Newton algorithm This should aid convergence and avoid possible overflow vmax Real maximum voltage allowed during iterations of the Newton algorithm vmin Real minimum voltage allowed during iterations of the Newton algorithm damping Boolean false apply a version of the damped Newton method ggroundstepping Boolean true turn on off the gground stepping continuation algorithm It acts as fall back method only when the conventional transient analysis fails due to singular MNA matrix gminstepping Boolean true turn on off the gmin stepping continuation algorithm It acts as fall back method only when the conventional transient analysis could fail pseudotransient Boolean true turn on off the pseudo transient continuation algorithm It acts as fall back method only when the conventional transient analysis could fail cmin Boolean true if true parasitic capacitors of suitable values are inserted between each circuit node and ground when the simulator is in trouble to compute the solution This can aid integration algorithms However attention must be payed since insertion of these capacitors can yield some ci
136. evice 91 pcgsl Cross term dependence of cgsl pcgdl Cross term dependence of cgdl pckappas Cross term dependence of ckappas pckappad Cross term dependence of ckappad pef Cross term dependence of cf pclc Cross term dependence of clc pcle Cross term dependence of cle palphaO Cross term dependence of alphaO palphai Cross term dependence of alphal pbeta0O Cross term dependence of beta0 pagidl Cross term dependence of agidl pbgidl Cross term dependence of bgidl pcgidl Cross term dependence of cgidl pegidl Cross term dependence of egidl paige Cross term dependence of aigc pbigc Cross term dependence of bigc pcigc Cross term dependence of cigc paigsd Cross term dependence of aigsd pbigsd Cross term dependence of bigsd pcigsd Cross term dependence of cigsd paigbacc Cross term dependence of aigbacc pbigbacc Cross term dependence of bigbacc pcigbacc Cross term dependence of cigbacc paigbinv Cross term dependence of aigbinv pbigbinv Cross term dependence of bigbinv pcigbinv Cross term dependence of cigbinv pnigce Cross term dependence of nigc pnigbinv Cross term dependence of nigbinv pnigbacc Cross term dependence of nigbacc pntox Cross term dependence of ntox peigbinv Cross term dependence for eigbinv ppigcd Cross term dependence for pigcd ppoxedge Cross term dependence for poxedge pvfbcv Cross term dependence of vfbcv pvfb Cross term dependence of vfb pacde Cross term dependence of acde pmoin Cross term dependence of moin pnoff Cross t
137. f circuit nodes belonging to the loop and 8 is the cardinality of the cut that opens the loop Branches 2176 77 L 33 where 2176 refers to the circuit node number is the histogram 77 is the number of branches L tells that the node belongs to s loop another possible character is S telling that the node belongs to a sub circuit and finally 33 shows the number that identifies the loop or sub circuit respectively Note that sizes less than a lower threshold are not displayed maxiters Integer 100 maximum number of iterations print Boolean false or list print operating point or a subset of node voltages and branch currents as specified in the given list Regular expressions can be used to specify the items of the list printnodes Boolen false print external internal node names index map This parameter serves mainly for debug purposes instance String device whose parameter value is to be swept 106 Chapter 4 Reference opt boolean false tell that a simulator options has to be swept param String name of parameter to be swept center Real center of sweep dec Integer points per decade lin Integer 50 0 number of points linear sweep log Integer 50 0 number of points log sweep span Real 0 0 sweep limit span start Real 0 0 start sweep limit step Real step size linear sweep stop Real stop sweep limit resetsweep Boolean true
138. f the round off error This is accomplished by reordering the rows and the columns of the Jacobian matrix Reordering requires the identification of the pivoting entries and submatrix reduction by meas of row column linear combinations This is an extremely cpu time consuming task that often takes a very large portion of the total simulation time mainly in large circuits Experience shows that if the Jacobian matrix is LU factorised by keeping a fixed reordering schema the total simulation time reduces of about 6 times The main drawback is that the growth of the round off error is badly controlled and can lead to a failure of the Newton method Once more experience shows that the Newton algorithm by its intrinsic iterative nature is almost always able to dump the effects of the round off errors but still there can be circuits yielding a very bad conditioned Jacobian matrix that can cause convergence problems stab Integer 0 the stability analysis of the computed equilibrium point is performed The pole with the largest real value is determined if it is positive the circuit is unstable Two different solvers can be chosen 0 no stability analysis is performed 1 a direct solver is employed 2 an iterative solver is employed nport Boolean false consider devices mainly mosfets as nports i e their internal nodes are dropped during the solution of the circuit The solution at internal nodes are computed 4 6 Basic analyses 107 thro
139. f the sweeping analysis If set to false parameters are left at values assumed at the end of the sweeping analysis center Real center of sweep dec Integer points per decade lin Integer 50 0 number of points linear sweep log Integer 50 0 number of points log sweep span Real 0 0 sweep limit span start Real 0 0 start sweep limit step Real step size linear sweep stop Real stop sweep limit instance String instance to which parameter sweeping is applied model String model to which parameter sweeping is spplied opt Boolean false tells that sweeping applyies to a simulator option specified through the lt option gt card mioparam String name of the parameter of the model instance or simulator option that has to be swept variable String name of the variable to be swept that has been specified in the body of a lt control gt analysis or in the body of the lt sweep gt analysis itself param String name of the parameter specified through the keyword parameter in the circuit netlist to be swept Control block The control analysis interprets a language similar to that of the octave environment It allows to perform elaborations of the simulation results to alter instance and model parameters and to perform analyses In other words it gives the possibility to implement algorithms that drive the simulation flow The interpreted language allows the definition of proced
140. fficient cjsw Side junction capacitance per area mjsw Side grading coefficient neff Total channel charge coefficient lambda Channel length modulation 1 Default length of the mos model is Bulk junction saturation current js Bulk juncction saturation current density tox Real 1 0e 7 oxide thickness 1d Lateral diffusion uo Surface mobility fc Forward bias juction fit parm nsub Substrate doping tpg Gate type nss Surface state density delta Width effect on threshold vmax Maximum carrier drift velocity xj Metallurgical junction depth nfs Fast surface state density delta Width effect on threshold tnom Parameter measurement temperature Celsius vt0O Threshold voltage vto Threshold voltage w Default width of the mos model uexp Critical field exponent in mobility degradation ucrit Critial field for mobility degradation rsh Sheet resistance rd Drain ohmic resistance default 10 mOhm rs Source ohmic resistance default 10 mOhm 4 4 6 4 4 semiconductor device 65 imax Real 100 0e 3 when current is above this value the diode is considered in the melting region and its characteristic is linearized vbox Real 1 0e9 tox oxide breakdown voltage kf Real 0 0 flicker noise coefficient af Real 1 0 flicker noise exponent MOSFET level 3 device The mosfet transistor model is the level three of Berkeley SPICE3 ELEMENT DEFINITION Name drain gate source bulk thermal ModelName lt parameter value
141. flicker noise coefficient Linear Two Terminal Capacitor The capacitor is tied between nl and n2 C is the capacitance in farads and is a modifiable parameter ELEMENT DEFINITION Name nl n2 ModelName lt parameter value gt ELEMENT PARAMETERS c Real DefaultCapacitace capacitance value m Integer 1 number of identical capacitors connected in parallel w Real capacitor width 1 Real capacitor length scale Real 1 0 capacitor scale factor 4 2 3 44 Chapter 4 Reference qv String the algebraic function that describes the q v charge voltage characteristic of the capacitor cv String the algebraic function that describes the c v capacitance voltage characteristic of the capacitor icon Initial capacitor voltage It can be used as initial condition by time domain algorithms like lt tran gt lt shooting gt lt envelope gt and lt mtshooting gt ic Initial capacitor voltage It can be used as initial condition by time domain algorithms like lt tran gt lt shooting gt lt envelope gt and lt mtshooting gt tci Real 0 0 first order temperature coefficient tc2 Real 0 0 second order temperature coefficient MODEL DEFINITION model ModelName capacitor lt parameter value gt MODEL PARAMETERS c Real DefaultCapacitace capacitance value c1 c1 Real 0 0 the coefficients describing the characteristic of a polynomial capacitor The charge Q is expresse
142. global one is selected If the global one does not exist an error is displayed There can be more than one of this parameter according to the number of input ports func The function that describes the characteristic of the CCVS The controlling currents can be accessed through the function i px where x is an integer number that refers to the input port For example Ic1 o1 02 il i2 cccs func i p1 defines a CCVS with the output port connected between the ol and 02 nodes and driven by one input port a short circuit connected between the il and i2 nodes The i p1 function refers to the first and unique driving port There is an other mechanism to access driving currents The circuit element whose current is chosen as driving is identified by the sensedev parameter In this case in the expression the current can be accessed through the function i v1 where v1 is the name on the instance used as argument in the sensedev parameter for example in the circuit card Icl o1 02 cccs sensedev v1 func i v1 the current of v1 is accessed through the function i v1 digital Itis a digital macromodel It is supposed that the output varied sharply between two 4 3 5 54 Chapter 4 Reference different and constant voltage levels trtime This option has meaning only if the lt digital gt option is set true defauls Ins Digital devices make only discrete variations of
143. gnd resistor r rmc MonteCarlo analysis control begin count 0 samples 100 0 Mc mc samples 100 begin De dc print no annotate 0 count count 1 samples count get Dc x end printf n printf n printf mean v x e std e n mean samples std samples printf n printf n endcontrol The voltage at the x output node is half that of the independent voltage source v1 Value of resistors are equal to the rme parameter In turn the rme parameter is equal to the rx 10 one Any time a parameter is followed by the lt lot process gt or lt dev process gt keywords a statistical process is assign to the parameter Therefore the above example rmc is linked to a lt dev gaussian gt process with variance equal to the value of the std 0 5 parameter The control analysis named MonteCarlo has a body that implements an interpreted language similar to that of Matlab Octave environments In particular the count 0 statement initialises the count variable and the samples 100 0 initialises to 0 the samples array The mc MonteCarlo analysis is executed in the control body that is the random number generators associated to parameters are exercised at each iteration of the MonteCarlo analysis the related body is execut
144. gt ELEMENT PARAMETERS 1 Real 2 0e 6 channel length w Real 2 0e 6 channel width ad Drain region area as Source region area pd Drain perimeter ps Source perimeter nrd Drain diffusion sheet resistance nrs Source diffusion sheet resistance region Operation region it is used to compute the guess to star the Newton method during the computation of the DC operating point Allowed values are region 0 off region 1 on in the first case the Vgs voltage is set equal to Vth and Vds 2 Vth in the second case Vgs 0 Default value of parameter is region 1 temp Device junction temperature MODEL DEFINITION model ModelName mos3 lt parameter value gt MODEL PARAMETERS cbd B D junction capacitance cbs B S junction capacitance gamma Bulk threshold parameter kp Transconductance parameter nmos Nmos type transistor phi Surface potential pmos Pmos type transistor imax Real 100 0e 3 when current is above this value the diode is considered in the melting region and its characteristic is linearized is Bulk junction saturation current pb Bulk junction potential cgso Gate source overlap cap cgdo Gate drain overlap cap cgbo Gate bulk overlap cap cj Bottom junction capacitance per area mj Bottom grading coefficient cjsw Side junction capacitance per area 4 4 7 66 Chapter 4 Reference mjsw Side grading coefficient 1 Default length of the mos model is Bulk junction satura
145. h current corner for forward beta ise B E leakage saturation current jle B E leakage saturation current ne B E leakage emission coef nle B E leakage emission coef br Reverse beta nr Reverse emission coefficient var Reverse early voltage vbr Reverse early voltage ikr High current corner for reverse beta jbr High current corner for reverse beta isc B C leakage saturation current jic B C leakage saturation current nc B C leakage emission coef nlc B C leakage emission coef rb Zero bias base resistance default 1 0e 3 irb Current for base resistance midpoint jrb Current for base resistance midpoint rbm Minimum base resistance for high currents re Emitter resistance default 1 0e 3 rc Collector resistance default 1 0e 3 imax The currents through the BE and BC junctions exponentially depend on the Vbe and Vbc voltages The imax parameter defines the maximum Vbe and Vbc voltages above which the exponential function is linearized Linearization is introduced to avoid numerical overflow The value of this parameter can be chosen close to the maximum allowed currents of the BJT The default value of this parameter is 1A cje Real 0 0 base emitter zero bias junction capacitance vje B E built in junction potential pe B E built in junction potential mje B E junction exponent me B E junction exponent cjc B C zero bias junction capacitance vjc B C built in junction potential pe B C built in junction potential mjc B
146. h the same name exist the local one is chosen There can be more than one of this parameter according to the number of input ports probe String the name of the instance of the sensing device The current through this device is used to control the CCCS If the instance of the CCCS belongs to a subcircuit the sensing device is searched locally This means that is a local and global devices with the same name exist the local one is chosen There can be more than one of this parameter according to the number of input ports gbsensedev String the global name of the instance of the sensing device The current through this device is used to control the CCCS Only global devices are considered devices at the top level of the netlist i e those not belonging to subcircuits are referred to as global If a global and a local device exist the global one is selected If the global one does not exist an error is displayed There can be more than one of this parameter according to the number of input ports func The function that describes the characteristic of the CCCS The controlling currents can be accessed through the function i px where x is an integer number that refers to the input port For example Ic1 o1 02 il i2 cccs func i p1 defines a CCVS with the output port connected between the o1 and 02 nodes and driven by one input port a short circuit connected between the il and i2 nodes The i p1 f
147. hat handles the arguments passed to the the macromodel Linear Current Controlled Current Source Positive current exits the source node and enters the sink node There are two ways for sensing current by specifing the sensing nodes or by giving the instance of a sensing device In the first case a short circuit is placed between the two given nodes Sense current enters sense node and leaves sense node In the second case only the source and sink nodes of the current controlled current source are given and a sense device for example a voltage source is specified Its branch current drives the current controlled current source ELEMENT DEFINITION Name sink source lt a lot of terminals gt cecs lt parameter value gt ELEMENT PARAMETERS 4 3 4 52 Chapter 4 Reference gaint Real 0 0 current gain of each driving port of the cccs since there can be more than one input port there must be as much of these parameters as the input ports or sensing devices gain Real 0 0 current gain of each driving port of the cccs since there can be more than one input port there must be as much of these parameters as the input ports or sensing devices delay Time delay sensedev String the name of the instance of the sensing device The current through this device is used to control the CCCS If the instance of the CCCS belongs to a subcircuit the sensing device is searched locally This means that is a local and global devices wit
148. he current in the generator as a variable providing its value in the output data file Initial Conditions and simulation of the Van der Pol oscillator Let us finally consider the first analysis line we will consider the second one later TranO tran tstop 100 uic 1 As always the first word Tran0 represent a unique name for the analysis while the second is the analysis type in our case the keyword tran asserting that this is a transient aka time domain analysis Option tstop is mandatory it represents the time at which simulation must stop In our case since this is a rather slow oscillator the rather large tstop 100s is used The second option uic 1 stands for use mode 1 for Initial Conditionss ICs Three different IC modes are allowed uic 0 is equivalent to no ICs i e perform a DC analysis to set ICs uic 1 use ICs given for all capacitors and inductors and set to 0 those that are not defined i e 0 voltage on capacitors and 0 current in inductors uic 2 use ICs given for all capacitors and inductors and assume that all capacitors and inductors whose initial conditions have not been specified are initially considered equivalent to open and short circuits respectively Initial conditions are set on each capacitor and inductor line with option icon value In our case the IC has been set for the capacitor while the inductor has been left unspecified since mode is set this means that the initial cu
149. he exact amplitude of the probe that cases the branch current of the voltage probe or the branch voltage across the current probe to be 0 This solution is then continued by sweeping the amplitude or the frequency of the voltage probe untill the modulus of the branch current of the voltage probe or the branch voltage across the current probe is 0 When this happens a solution of the steady state problem has been found Note that the insertion of the voltage probe can be automatic that is the method automatically determines the nodes to which connect the voltage probe The continuation with no hybrid newton acceleration acts as the continuation method by does not apply the hybrid newton to try to accelerate convergence to a steady state solution In fact hybrid newton does not converge with some circuits and thus wasts cpu time The newton solver is based on the insertion of a voltage probe or a current probe in the circuit It adopts the Newton iterative method to find the 0 value of the branch current of the voltage probe or of the branch voltage of the current probe The unkowns are the magnitude and the frequency of the probe In general convergence is achieved if the Newton method is started from a suitable initial condition It is suggested to perform a pole zero analysis and to employ the frequency of the unstable pole as initial guess The conventional solver adds the working frequency of the autonmous circuit among the probl
150. he transient analysis related to the lt pnoise gt analysis Before performing a lt pnoise gt analysis the circuit is linearised along one working period The solution along this working period is determined by a transient analysis The lt tmax gt parameter upper limits the maximum integration time step that can be used by the transient analysis Advanced frequency domain analyses Pole Zero Analysis This analysis linearises the circuit about the DC operating point and determines the poles and zeros of the linear circuit In particular to compute zeros a transfer function must be defined by specifying the input source and the output voltage or current ANALYSIS DEFINITION Name pz lt parameter value gt ANALYSIS PARAMETERS annotate Integer default is global annotate level degree of annotation Values from 1 to 7 turn on display of generic information about the analysis run Values above 7 are bitwise and turn on specific printouts 0x08 print MNA matrix 0x10 dump MNA matrix in the analysis_name m file statematrix Boolean false dump pseudo state matrix only with direct solver thermalnet Boolean false use the thermal network hzout Boolean true Output pole zero values as Hz or as equivalent eigenvalues solver Integer 2 type of solver solvers to be used by the analysis Possible values are 1 iterative solver 2 direct method 3 iterative solver followed by the direct one In general the iterative s
151. his option has meaning only when the fundamental matrix is computed Saltation matrices are exploited to construct a correct fudamental matrix of circuits characterised by discontinuities in the vector field Discontinuities in the vector field can be also introduced by sharp varying functions modeling circuit elements such as for example the y tanh a x function with the a gain too large The solver automatically determines possible discontinuities due to numerical problems introduced by too sharp nonlinear functions and automatically introduce saltation matrices The solver introduces also saltation matrices due to dscontinuities caused by commutations of pseudo analog to digital and digital to analog converters during simulations of mixed digital analog circuits and due to behavioural elements tagged as digital lyapunov Boolean false compute the Lyapunov exponents and the Kaplan York dimension of the tran circuit trajectory It is possible to chose different policies concerning the time instant at which monodromy is reset and QR computed Time based after a given time or Sample based after a fixed number of time points Sample based is the default behaviour if parameter lyapstep is not specified lyapstep Real Force Time Step for the computation of Lyapunov exponents and resetting monodromy to identity lyapsamp Integer 100 Number of time samples that force computation of QR and resetting monodromy matrix
152. ial njd Drain junction emission coefficient xtid Drainjunction current temperature exponent mjd Drain bottom junction capacitance grading coefficient pbswd Drain sidewall junction capacitance built in potential mjswd Drain sidewall junction capacitance grading coefficient pbswgd Drain gate side sidewall junction capacitance built in potential mjswgd Drain gate side sidewall junction capacitance grading coefficient cjd Drain bottom junction capacitance per unit area cjswd Drain sidewall junction capacitance per unit periphery cjswgd Drain gate side sidewall junction capacitance per unit width vfbcv Flat Band Voltage parameter for capmod 0 only vfb Flat Band Voltage tpb Temperature coefficient of pb tcj Temperature coefficient of cj tpbsw Temperature coefficient of pbsw tcjsw Temperature coefficient of cjsw tpbswg Temperature coefficient of pbswg tcjswg Temperature coefficient of cjswg acde Exponential coefficient for finite charge thickness moin Coefficient for gate bias dependent surface potential noff C V turn on off parameter voffcv C V lateral shift parameter dmcg Distance of Mid Contact to Gate edge dmci Distance of Mid Contact to Isolation dmdg Distance of Mid Diffusion to Gate edge 4 4 semiconductor device 8 dmcgt Distance of Mid Contact to Gate edge in Test structures xgw Distance from gate contact center to device edge xgl Variation in Ldrawn rshg Gate sheet resistance ngcon Number of gate contacts xr
153. idered connected to ground The format is lt probel pos neg gt vprobe2magi Real 0 0 the magnitude of the voltage probe acting of the second funda mental vprobe2phi Real 0 0 the phase of the voltage probe acting on the second fundamental vprobe2ri Real 0 0 the value of the resistor connected in series to the voltage probe acting on the second fundamental vprobe2max Real the maximum value of the magnitude of the voltage probe employed during the sweep to find the steady state solution vprobe2freqit Integer 0 the index of the first fundamental or harmonic that beats with the frequencies and harmonics of the second probe related to the second fundamental Small Signal S Parameter Analysis This analysis linearizes the circuit about the DC operating point and performs a small signal analysis If desired the analysis will be performed repeated while sweeping some independent 4 9 5 4 9 Advanced frequency domain analyses 143 variable The variable may be either frequency or some device instance parameter Not all device instance parameters are modifiable and some of those that will change the DC operating point when modified If changing a parameter affects the DC operating point it will be recomputed on each step The modified parameter is reset to its original value after the analysis ANALYSIS DEFINITION Name sp lt parameter value gt ANALYSIS PARAMETERS annotate Integer defau
154. iffusion area 1s Real 1 0e 06 length of the sidewall of the diffusion area ab which is not under the gate 1g Real 1 0e 06 length of the sidewall of the diffusion area ab which is under the gate region Boolean 0 supposed DC operating region O off 1 on m Integer 1 the number of identical devices that are connected in parallel MODEL DEFINITION model ModelName juncap lt parameter value gt MODEL PARAMETERS ntype Boolean false the juncap is a ntype ptype Boolean true the juncap is a ptype type Tag type of juncap Admissible types are p and n tnom Real temperature at which the parameters have been determined tref Real temperature at which the parameters have been determined tr Real temperature at which the parameters have been determined vr VR dta DTA jsgbr JSGBR jsdbr JSDBR jsgsr JSGSR jsdsr JSDSR jsggr JSGGR jsdgr JSDGR nb NB ns NS ng NG vb VB 4 5 45 1 4 5 2 4 5 misc device 95 cjbr CJBR cjsr CJSR cjgr CJGR vdbr VDBR vdsr VDSR vdgr VDGR pb PB ps PS pg PG misc device analog to digital converter This element implements a pseudo digital to analog converter that links the digital circuit to the electrical one ELEMENT DEFINITION Name pos neg d2a lt parameter value gt ELEMENT PARAMETERS dignet String the hierarchy path inside the design that identifies the net to which the converter is connected For example Mod U1 e s
155. in the right panel Steady state time domain analysis shooting There are many situation where the designer is interested in finding the periodic steady state solution of a circuit While a first and naive approach could be that of performing a long time domain simulation waiting for the solution found to settle in a stable situation this could lead to 2 5 1 30 Chapter 2 A simple PAN walk through 3 5 3 57 2 5 2 57 o o gt gt 1 5 1 5f Env0O Env0O Trano Trano 0 5 0 5 0 0 19 time time Figure 2 12 Envelope analysis for the bubba oscillator followed by a transient analysis in the left panel the full simulation time is shown The envelope anaysis as expected fully simulates the oscillator only in short time windows In the right one a detail of the final time of the envelope analysis and start of the transient is shown very long simulation times for certain classes of circuits such as high Q oscillators The shooting method overcomes this problem by solving a boundary value problem that forces a periodicity condition by finding an initial condition x9 such that the circuit after one period is back at xo The circuit can be autonomous or non autonomous simply meaning in the second case that the state equations of the circuit have explicit dependence on time or from a practical point of view that there are independent periodic voltage or current generators in the ci
156. in these cases the LU factorisation method is chosen there are convergence difficulties of the shooting method since the residue can not be suitably reduced 2 iterative solver based on hybrid Newton and saltation matrices It is suited to the computation of the steady state solution of problems where electrical variables shows discontinuities of the first kind Note that mixed analog digital circuits fall in this class save String the path of the file into which the solution is saved bz2 format is adopted The solution stored in this file can be later loaded by the simulator and used as initial condition from which the shooting algorithm is started load String the path of the file from which a previously saved solution is read The file format can be both ascii and bz2 and it is automatically determined during loading This solution is used as initial guess to the shooting algorithm Steady state computation through the multi tone shooting method Experimenal analysis use at your own risk What it does will be detailed in a future release ANALYSIS DEFINITION Name mtshooting lt parameter value gt ANALYSIS PARAMETERS annotate Degree of annotation It is an integer number from 1 to 7 that turns on displaying of generic information about the analysis run Values above 7 are bitwise and turn on specific 4 8 Advanced time domain analyses 133 printouts In particular 0x08 print sensitivity matrix and RHS at various
157. inductors for example If a capacitor is connected between nodes k and j two Floquet multipliers are computed one for node voltage k and the other for node voltage j The vB t eigenfunctions refer to the product of the v t ones with the B t function The B t function represent the sensitivity of pseudo state variables for example node voltages with respect to current injected into the node the same comment can be aplied to branch currents For more theoretical details see M Farkas Periodic Motions Springer Verlag 1994 A Demir A Mehrotra and J Roychowdhury Phase Noise in Oscillators A Unified Theory and Numerical Methods for Characterization IEEE Trans on CAS I Vol 47 No 5 May 2000 pp 655 674 A Brambilla P Maffezzoni G Storti Gajani Computation of Period Sensitivity Functions for the Simulation of Phase Noise in Oscillators IEEE Trans on CAS I Vol 52 No 4 April 2005 Waveforms that can be stored in memory and made available to the control analysis are 1 Node voltage and branch current waveforms can be retrieved through the statement get ShootName node time is stored in the time vector that can be retrieved through get ShootName time 2 Node voltage and branch current spectra can be retrieved through the statement get ShootName node spectrum The frequency samples are stored in the freq vector It can be retrived th
158. ine describing the macromodel is called with the following ordered arguments lt input gt int an index that uniquely identifies the instance of the CCVS This index starts from 0 and it is incremented by 1 each time a new instance is found in the netlist lt input gt int 4 3 Basic two port device 53 number of ports of this CCVS lt input gt double the array of the controlling currents lt input gt int number of state variable lt input gt double the two vectors of state variable values at the current n time point index 0 and at the previous n 1 time point index 1 In other words the array can be indexed as array 0 StateVarNum current time point and array 1 StateVarNum previous time point lt output gt double the array of the transresistances the meaning is the same as that of the lt gain gt parameter lt output gt double the value of the output port branch voltage lt input gt double the value of simulation time It has meaning only in time domain analyses The voltage of the controlled port can also dependent on the derivative of the currents through the controlling ports In this case the CCVS acts as a derivator In time domain analyses the Gear method of order 1 or 2 is employed to numerically compute derivatives ELEMENT DEFINITION Name sink source lt a lot of terminals gt ModelName lt parameter value gt ELEMENT PARAMETERS gaint Real 0 0 gain of each drivi
159. ines a list on nodes or branch currents that will be saved into the output file during simulation Other computed electrical variables are not saved devcurr Boolean false the currents of devices for example the drain current of mosfets the anode cathode current of diodes and currents through capacitors are stored in the simulation output file thermalnet Use the thermal network partitioner Boolean false this option turns on off the circuit partitioner supernode Boolean false introduce custsets formed by independent voltage sources These cutsets are known as super nodes Super nodes avoid to add the current through the independent voltage source among the unkowns of the problem and thus reduce the size of the problem increasing efficiency subeqns Integer 1000 number of equation to feed each thread circuit partitioning threads Boolean flase turn on off the parallel analysis by using multiple threads 4 6 Basic analyses 11 method Select the integration method 1 trapezoidal default 2 Gear method of various order order Set the order of the Gear integration method default value is 2 It has no meaning if other integration methods are employed acntrl Integer 2 select the accuracy control method 1 relative energy balance this method computes the electrical energy variation of each capacitor and inductor along each integration interval and the corresponding electrical work delivered by the resistive one port
160. integer number values from 1 to 7 turn on displaying of generic informations about the analysis run start Start sweep limit stop Stop sweep limit center Center of sweep span Sweep limit span step Step size linear sweep lin Number of points linear sweep dec Points per decade log Number of points log sweep freq Real the frequency value in single frequency analysis This parameter must not be 144 Chapter 4 Reference specified contemporary to the frequency sweeping command onodes List the list containing the pair of nodes the ground node can be omitted at which the noise is computed The format is lt onodes node1 node2 gt Sth teas POF I ha PURS ii UAR iR F i l ap HAN jpe sy rots ma i Ekiti R mE 4 r meai At ANEAN E atil N vi Gs His ah mate Ti gesi i ti L _ Books CDK87 L O Chua C A Desoer and E S Kuh Linear and Nonlinear Circuits New York McGraw Hill Editions 1987 cited on page 12 Far94 Mikl s Farkas Periodic motions New York NY USA Springer Verlag New York Inc 1994 ISBN 0 387 94204 1 cited on page 33 Kuz95 Y A Kuznetsov Elements of Applied Bifurcation Theory Springer Verlag 1995 cited on page 33 Rhe95 R W Rhea Oscillator Design and Computer Simulation New York McGraw Hill 1995 cited on page 30 VS83 J Vlach and K Singhal Computer Methods for Circuit Analysis and Design Van Nostrand Reinhold Company 1983 cited
161. inued as far as we know The command line that extracts from the schematic file bubba2 sch the netlist is prompt gt gnetlist g spice sdb o bubba2 pan 0 include_mode 0 sort mode bubba2 pan 28 Chapter 2 A simple PAN walk through r 7m bubba2 sch ggchem File Edit Buffer View Page Add Hierarchy Attributes Options Help g g Tl jk A1 SPICE include File TLV247xA sub 360k 10k n5 sin_out EEEH HHR A2 SPICE directive Cp T Tran0 tran tstop 210m tmax 0 0 1m restart no A3 SPICE directive Envo envelope tstop 200m period 1 5m autonomous yes uic 1 Pick Repeat none Menu Cancel Grid 100 100 Select Mode Figure 2 11 The Bubba oscillator in the geda tools gschem window Netlist 2 4 1 the line above instructs gnetlist to extract the netlist to file bubba2 pan Options 0 include_mode 0 sort mode are for the backend and request the netlister to include external models without explicitly embedding them in the final netlist and to write a netlist sorted by instance name The resulting netlist is shown in 2 4 17 As it can be seen the SPICE directive pseudocomponent has been used to include commands to execute simulations another viable solution possibly more efficient in many cases is to include an external file containing all simulator directives and options or to directly edit the netlist file to add simulator directives as desired The final netlist is in spice syntax but PAN aut
162. ipated by independent sources is lt 2 1000 mW gt Convergence achieved in lt 2 gt iterations Current solution of lt Dc0 gt DC analysis V n1 2 0000 V V n2 10 050 mV V n3 2 0201 V V n4 1 92001 V LEa 2 0101 m CPU time for this analysis was 0 00 seconds Elapsed time was 0 00 Physical memory usage was 296 kBytes Resource usage CPU user time 0 00 seconds CPU system time 0 00 seconds Execution date Sat May 31 15 57 54 2014 PAN reads the netlist and after a number of messages asserting execution status prints a circuit structure summary In our case 4 nodes and 5 equations node count does not consider the reference node The equations are 5 since the vsource is a bad branch i e not voltage controlled and thus requires the introduction of an additional equation The device summary section reports the number of device instances found and is quite interesting when large circuits are simulated After these preliminary steps analysis are performed one at a time in the order they are encountered In our case we have one single DC analysis called DcO After a few messages about gground resistors and initial guess for the Newton algorithm the meaning of these 2 2 The DC analysis 15 messages will be discussed later DC simulations is performed Total power absorbed by independent sources is always output to log and screen current solution i e the 4 node voltages and the current in
163. is 2 local method Select the integration method 1 trapezoidal default 2 Gear method of various order order Set the order of the Gear integration method default value is 2 It has no meaning if other integration methods are employed hgfactor Set the growing factor of the envelope step H At each iteration if the local truncation 4 8 Advanced time domain analyses 125 error is below the threshold set through the lt ereltol gt and lt eabstol gt parameters the envelope step is increased by a certain ammount The maximum allowed increase is set by the estimated local truncation error but the growing factor can be set by the user default is 2 maxh Maximum allowed value of the envelope integration step H as default H is unlimited corrector This parameter defines the type of corrector its value must lie from 0 5 pure trapezoidal method to 1 0 pure Implicit Euler method default timesteps List the list containing the integer multiples of the perio of the envelope waveform at which a time domain analysis must be performed In case of autonomous circuits the period varies and is automatically determined by the envelope analysis In case of non autonomous circuits the value of period is given by the user and does not vary during the envelope analysis In both cases the envelope integration time step H is adjusted to eaxcly meet the integer multiples specified in the list timeinstants List the li
164. itting parameter for source diode breakdown current bvd Drain diode breakdown voltage bvs Source diode breakdown voltage jtss Source bottom trap assisted saturation current density jtsd Drain bottom trap assisted saturation current density jtssws Source STI sidewall trap assisted saturation current density jtsswd Drain STI sidewall trap assisted saturation current density jtsswgs Source gate edge sidewall trap assisted saturation current density jtsswgd Drain gate edge sidewall trap assisted saturation current density njts Non ideality factor for bottom junction njtssw Non ideality factor for STI sidewall junction njtsswg Non ideality factor for gate edge sidewall junction 4 4 semiconductor device 83 xtss Power dependence of JTSS on temperature xtsd Power dependence of JTSD on temperature xtssws Power dependence of JTSSWS on temperature xtsswd Power dependence of JTSSWD on temperature xtsswgs Power dependence of JTSSWGS on temperature xtsswgd Power dependence of JTSSWGD on temperature tnjts Temperature coefficient for NJTS tnjtssw Temperature coefficient for NJTSSW tnjtsswg Temperature coefficient for NITSSWG vtss Source bottom trap assisted voltage dependent parameter vtsd Drain bottom trap assisted voltage dependent parameter vtssws Source STI sidewall trap assisted voltage dependent parameter vtsswd Drain STI sidewall trap assisted voltage dependent parameter vtsswgs Source gate edge sidewall trap assisted voltage dependent paramete
165. lly determined during loading This solution is used as initial guess to the analysis save String the path of the file into which the solution is saved bz2 format is adopted The solution stored in this file can be later loaded by the simulator and used as initial condition from which the analysis is started fund Real the fundamental frequency at which the circuit works If the considered circuit is autonomous this value gives an initial estimate of the final working frequency computed by the harmonic balance analysis period Real the period at which the circuit works If the considered circuit is autonomous this value gives an initial estimate of the final working period computed by the harmonic balance analysis harms Integer 2 the number of harmonics employed by the harmonic analysis thresh Real 1 0e 15 during the harmonic analysis each nonlinear or time varying element is characterised by a spectrum that contributes off diagonal entries to the block organised MNA matrix The numerical effort to LU factorise this block matrix at each iteration of the Newton algorithm depends on the bandwidth of each block sub matrix In general the magnitude of entries lower as they are far from the main diagonal high order harmonics This parameter defines a relative threshold with respect to the largest magnitude of the spectrum components of each device below which the entries can be neglected A suitable 4 9 Advanced frequency dom
166. ls with loops of voltages source and capacitors even though consistent initial conditions are given since the loop current is undertermined The same applies to inductor cut sets 2 it is assumed that all the capacitors and inductors whose initial conditions have not been specified are open circuit and short circuit respectively In this case suitable voltage source and capacitor loops orcurrent source and inductor cut sets do not cause any problem vabstol Absolute voltage tolerance employed by the Newton algorithm at each time point of the time domain analysis vreltol Relative voltage tolerance employed by the Newton algorithm at each time point of the time domain analysis saman Boolean false turns on off the samanskii method Periodic Small Signal Analysis The pac analysis computes transfer functions of circuits that admit a periodic solution These circuits can be both autonomous oscillators and non autonomous The pac analysis linearises the circuit along the limit cycle defined by the periodic large signal solution Non linear elements are transformed in time varying linear ones that thus exibit frequency translations The pac analysis must be preceeded by a shooting analysis that determined the limit cycle of the unperturbed circuit Then the effects of small signal perturbations assumed additive are computed by the pac analysis ANALYSIS DEFINITION Name pac lt parameter value gt ANALYSIS PARAMETERS annotate Integer
167. lt is global annotate level degree of annotation Values from 1 to 7 turn on display of generic information about the analysis run Values above 7 are bitwise and turn on specific printouts 0x08 print MNA matrix 0x10 dump MNA matrix in the analysis_name m file freq Real the frequency value when a device value parameter is swept or the frequency value in single frequency analysis This parameter must not be specified contemporary to the frequency sweeping command port Active port may be repeated file S parameters output file name start Start sweep limit stop Stop sweep limit center Center of sweep span Sweep limit span step Step size linear sweep lin Number of points linear sweep dec Points per decade log Number of points log sweep sweep Device whose parameter value is to be swept param Name of parameter to be swept Noise analysis of a periodic circuit Noise analysis computes the total noise at the specified port of a circuit Noise is computed by linearising the circuit in the operating point therefore a DC analysis is done before any noise analysis if a valid operating point is not available yet The thermal shot and fliker sources of noise in circuit elements are considered Note that since the circuit is linearised in the operating point currents of elements have constant values ANALYSIS DEFINITION Name noise lt parameter value gt ANALYSIS PARAMETERS annotate Degree of annotation It is an
168. ltage mag Small signal power dBm phase Small signal phase magi Power dBm of fundamental phase1 Phase of fundamental mag2 Power dBm of harmonic 2 phase2 Phase of harmonic 2 mag3 Power dBm of harmonic 3 phase3 Phase of harmonic 3 mag4 Power dBm of harmonic 4 phase4 Phase of harmonic 4 mag5 Power dBm of harmonic 5 phase5 Phase of harmonic 5 mag6 Power dBm of harmonic 6 phase6 Phase of harmonic 6 mag Power dBm of harmonic 7 phase7 Phase of harmonic 7 mag8 Power dBm of harmonic 8 phase8 Phase of harmonic 8 mag9 Power dBm of harmonic 9 phase9 Phase of harmonic 9 4 5 7 100 Chapter 4 Reference mag10 Power dBm of harmonic 10 phase10 Phase of harmonic 10 magit Power dBm of first fundamental phaseit Phase of first fundamental mag2t Power dBm of second fundamental phase2t Phase of second fundamental noise Real power spectral density of white noise dBm expressed as V7 Hz Note that if this parameter is given its value overwirtes the magnitude of the equivalent noise source derived from the given series resistance of the port In other words the series resistor is considered as noiseless Poly Silicon Thin Film Transistor Poly silicon thin film transistor Poly Si TFT model has 3 electrical terminals Poly Si TFT model has the feature of binning The binning equation is given by P PO P1 Leff Pw Weff Pp Leff Weff Only the PO parameters are listed and marked with BIN in desc
169. mber of time points employed to sample the large signal steady state solution and to define the size of the related variational model When not specified it is derived from the number of harmonics defined by the lt harms gt parameter type Integer 7 types of noise sources that are considered during the analysis Possible choices are 0x1 thermal noise 0x2 shot noise 0x4 flicker noise The value of this parameter can be the result of the bitwise or of types of noise sources decomp Boolean false when true the resulting component of the noise spectra along the ul t eigenfunction and those obtained by difference with it are computed These components have meaning only for autonomous circuits The component along the ul t 49 4 9 1 4 9 2 4 9 Advanced frequency domain analyses 137 eigenfunction is defined as phase noise and that obtained by subtracting this component from the complete result is defined as amplitude noise relnf Real ALU resolution dB this parameter defines the relative noise flore that can be introduced by the pnoise analysis Components of the transfer functions below this value are not considered solver Integer type of solver employed in the pnoise analysis available ones are 0 direct solver 1 iterative solver If the solver is not specified the same one employed in the related shooting analysis is adopted tmax Real maximum integration time step allowed during t
170. me Figure 3 4 Steady state trajectory of the PERO oscillator plotted vs time Note that since the shooting analysis is restarted from the previous transient one the time axis starts at 300 at T P i vr hs ii fT iri ud io Where and How downloading and installing The PAN circuit simulator binary can be downloaded from http brambilla ws dei polimi it The PAN binary requires a standard linux installation PAN developers use Debian based systems most recently Debian Mint and Ubuntu but PAN has been used without any pain with other Linux flavours including RedHat Gentoo Sabayon and several others A Linux live CD image including possibly not the latest version of the simulator and of the demo files is also available This allows quick evaluation without actually installing anything in your computer Using PAN is relatively simple and only requires knowledge of basic Linux terminal com mands and the use of any ASCII editor such as Vi As it is netlists are entered in ASCII format but as it will be shown in the sequel thanks to the high compatibility of PAN with well known simulator languages almost any standard netlister can be used as long as some adjustments to the command language are made A quick reference is available with pan s a list of keywords appears and further help on a specific topic can be obtained with the command pan s topic In the following sections a quick reference for all devices and a
171. med with the unique analysis identifier specified in the netlist file These files are in standard SPICE3 rawfile format and can thus be viewed using several visualization tools such as the open source tools gwave KJwave or commercial tools such as Synopsys sx or CustomWave The visualization tool iplot is also available at the PAN dowload site The tool is based on a rather old but still quite efficient set of libraries It is also possible to import rawfiles into Matlab or Octave using one of the several rawfile import filters easily available on the web The DC analysis The DC analysis computes the operating point of linear and nonlinear circuits The default behaviour of the simulator is to compute one single operating point even if the circuit admits more than one but analysis also implements methods to look for more than one operating point Like most methods implemented in PAN it is based on the method known as Modified Nodal Analysis MNA HRB75 this means that the default behaviour of the simulator is to compute only node voltages and current of circuit ports that are not voltage controlled these are often called bad branches An introduction to MNA can be found in many classic books see e g CDK87 VS83 Obviously if we wish to explicitly compute the value of a specific current of the circuit the simplest solution is to introduce a null independent voltage source in series with branch whose current we are interested i
172. mn with all entries equal to 0 If the second argument is omitted a column vector of length n is created Special variables anaerror stores the error code after the execution of an analysis This variable must be defined and initialised before its usage If the user does not define this variable or if the error is a fatal one the error is handled by the simulator Warnings and notices are discharded When an error occurs the lt anaerror gt variable is set equal to 1 A simple usage example follows anaerror 0 Alter alter instance R1 param r value 0 if anaerror gt 0 printf Alter error end The lt Alter gt analysis chances the value of the lt R1 gt resistor to 0 Since resistance can not be set to 0 an error is generated null this variable is used as argument of functions to specify that an optional argument possibly followed by other ones is not used ANALYSIS DEFINITION Name control lt parameter value gt ANALYSIS PARAMETERS annotate Integer default is global annotate level degree of annotation Values from 1 to 7 turn on display of generic information about the run 4 7 5 122 Chapter 4 Reference Monte Carlo Analysis The MonteCarlo analysis is introduce by means of an example A circuit divider composed of the r and r resistors is considered ground electrical gnd parameter rx 10 std 0 5 rmc rx dev gaussian std v nio gnd vsource vdc 10 ri nio x resistor r rmc r2 x
173. n A very simple example As a first example consider the circuit in Figure 2 1 and the corresponding Netlist 2 2 1 100Q 1kQ 00 2V ImA Figure 2 1 A simple example Netlist 2 2 1 Simple netlist File Simple_001 pan ground electrical 0 Analyses DcO de print yes 3 Circuit ri ni n2 resistor r 1k ci n2 0 capacitor c lu r2 n3 0 resistor r 2k 2 2 The DC analysis 13 r3 n3 n4 resistor r 100 tci n3 n2 n2 0 vccs gaini 0 2 vi ni O vsource vdc 2 ii n4 0 isource idc im Each device instance in this circuit is entered in its own line and correspondence of figure to netlist should be self evident The two port voltage controlled current source is defined by first giving respectively the sink and source pins followed by the sense nodes each pair of sense nodes requires a corresponding gain parameter Controlled sources can be described also in other ways please refer to the output of command prompt gt pan s vccs or the same command with vevs ccvs or cccs for more information The DC analysis command is on a single line DcO de print yes The DcO is a unique identifier more than one DC analysis with different options can be issued in the same file each must have its unique name Option print yes instructs the simulator to print the DC solution that has been found to output Default value of this parameter is no and in this case the operating point is
174. n O off 1 on temp Device junction themperature noisetype Integer 7 types of generated noise possible choices are 0x0 noiseless diode 0x1 thermal noise 0x2 shot noise 0x4 flicker noise The value of this parameter can be the result of the bitwise or of types of noise sources m Integer 1 number of devices connected in parallel MODEL DEFINITION model ModelName diode lt parameter value gt MODEL PARAMETERS is Saturation current js Saturation current jsw Real 0 0 sidewall saturation current isw Real 0 0 sidewall saturation current area Junction area in square meters rs Real 0 0 parassitic resistor connected in series to the diode that accounts teh resistance of contacts and of silicon that does not constitute the junction rsw Real 0 0 sidewall series resistance n Emission coefficient ik Real forward injection knee current ikf Real forward injection knee current ikp Real side wall forward injection knee current tt Real 0 0 transit time cjo Real 0 0 zero bias junction capacitance cjO Real 0 0 zero bias junction capacitance vj Junction potential pb Junction potential m Grading coefficient mj Grading coefficient cjsw Real 0 0 zero bias sidewall junction capacitance cjp Real 0 0 zero bias sidewall junction capacitance vjsw Sidewall Junction potential mjsw Sidewall Grading coefficient eg Real 1 11 band gap xti Real 3 0 saturation current
175. n our example these lines are included if symbol 22 Chapter 2 A simple PAN walk through STARTUP is defined and inclusion of these lines will activate a startup circuit Symbols can be defined inside the netlist using in our case with instruction define STARTUP or on the command line i e prompt gt pan vregu pan ad STARTUP where command line option aq stands for analog define Let us now simulate this voltage regulator without using the startup circuit We obtain the logfile shown only a small in part below Logfile 2 2 8 Logfile of the voltage regulator vregu log The computed equilibrium point is stable pole at 1 70e 38Hz Current solution of lt Dc1 gt DC analysis V vcc 20 000 V V a 20 000 V V b 19 668 V V 1 20 000 V V c 19 549 V V d 3 7595 mV V e 196 12 nV V f 215 68 nV V g 642 26 mV V o 3 2307 nV V out 17 068 mV I vcc i 0 0000 A v 0 1 800 V 1 10 0 Sweeping status v 0 2 5414 V RHS 16 713 pA RHS 1 1206 uA The computed equilibrium point is stable pole at 1 70e 38Hz V f 3 5502 V V g 19 964 V V o 3 4561 V V out 19 152 V I vec i 0 0000 A The number of found solutions is 5 In the full log file five solutions are shown and stability of each solution is computed by local linearization in the equilibria If the same circuit is simulated with STARTUP defined only one solution is found W
176. n the controlling voltage is greater than von if von gt voff and vice versa when the controlling voltage is less than von if von lt voff roff Real 1 0e6 off resistance The switch is considered off when the controlling voltage is less than voff if von gt voff and vice versa when the controlling voltage is greter than voff if von lt voff rin Real gmin input resistance of the controlling port von Real 1 0 on voltage threshold voff Real 1 0 off voltage threshold af Real 1 0 flicker noise exponent kf Real 0 0 flicker noise coefficient 4 6 4 6 1 4 6 Basic analyses 105 Basic analyses DC Analysis The DC analysis computes the operating point of linear and nonlinear circuits Furthermore it is possible to sweep one or more circuit parameters and compute the circuit operating point at each value of the parameters Computation is done through the Newton iterative method The DC analysis implements some continuation methods such as gmin stepping source stepping ggroung stepping and pseudo transient These methods should improve robustness and are called only when the conventional DC analysis failes The DC analysis supports also thermal networks in other words it is possible to carry out an electro thermal simulation where the self heating of devices is considered and the solution of the circuit accounts for the present values of device temperatures When dealing with large circuits they are partitioned to inc
177. nalyses available in PAN is provided In most cases these are identical to the inline quick reference Basic one port device Linear Two Terminal Resistor The resistor is tied between the nl and n2 nodes a possible third terminal is for a thermal connection and it has meaning only during electro thermal analyses During electro thermal simulations the self heating effects of the resistor are considered if its thermal node is connected to the thermal network The r parameter is the resistance in ohms it can be either positive or negative but it must not be zero Two types of resistors are admitted linear and piece wise linear The piece wise linear resistor is described by giving the coordinates of the extremes of the segments composing the characteristic in the voltage current plane The piece wise characteristic is continued adove and below the given extremes with a slope equal to that of the correspoding 42 Chapter 4 Reference last extreme Orizontal segments zero conducance are allowed but vertical ones are not To aid convergence of numerical algorithms a parasitic resistor of conductance lt gmin gt is connected in parallel to each pwl resistor ELEMENT DEFINITION Name nl n2 tl ModelName lt parameter value gt ELEMENT PARAMETERS r Real resistance value g Real conductance value m Integer 1 number of identical resistors connected in parallel func String the algebraic function that describe
178. nce of the VCVS This index starts from 0 and it is incremented by 1 each time a new instance is found in the netlist lt input gt int number of ports of this VCVS lt input gt double the array of the port voltages lt input gt int number of state variable lt input gt double the two vectors of state variable values at the current n time point index 0 and at the previous n 1 time point index 1 In other words the array can be indexed as array 0 StateVarNum current time point and array 1 StateVarNum previous time point lt output gt double the array of the transconductances the meaning is the same as that of the lt gain gt parameter lt output gt double the value of the branch voltage at the the output port lt input gt double the value of current simulation time It has meaning only in time domain analyses lt input gt double the value of simulation time step It has meaning only in time domain analyses The routine handling parameters of the macromodel is called with the arguments lt input gt int an index that uniquely identifies the instance of the VCVS lt input gt char parameter name lt input gt double parameter value ELEMENT DEFINITION Name v v lt a lot of terminals gt ModelName lt parameter value gt ELEMENT PARAMETERS gaint gain2 Real 0 0 gain of each driving port of the VCVS since there can be more than one input port there must be as much of
179. nd off the gmin stepping continuation method ggroundstepping Boolean flase turn on and off the gground stepping continuation method fund1 First fundamental frequency fund2 Second fundamental frequency harms Integer 1 largest considered harmonic of the first and second fundamental harms1 Integer 1 largest considered harmonic of the first fundamental harms2 Integer 1 largest considered harmonic of the second fundamental trunc Integer 0 the truncation schema used in DFT to define the components of the spectrum considered in the analysis Available choices are 0 square 1 diamond time Boolean 0 output time domain signals interval Real time domain output interval in seconds points Integer 200 number of time points in output waveforms vabstol Real 1 0e 6 Absolute voltage convergence criterion vreltol Real 1 0e 6 Relative voltage convergence criterion iabstol Real 1 0e 12 Absolute current convergence criterion ireltol Real 1 0e 6 Relative current convergence criterion autonomous Integer 0 solve an autonomous circuit The following different methods to find a steady state solution can be used 1 continuation 3 continuation with the insertion of only one probe 4 hybrid newton 5 conventional The continuation solver is bases on the insertion of a probe in the circuit It is composed of different phases In the first phase the magnitude of the probe is swept starting from 0 till a
180. ng the name of the routine that handles the arguments passed to the macromodel limiting String the name of the routine in the shared executable that does limiting of the electrical variables Limiting should be an aid to convergence of the Newton algorithm when strong nonlinearities are involved Independent Resistive Source A port is a resistive source used when calculating S parameters The port is tied between n and n R is the reference resistance It must not be zero or negative and has a default value of 50 ohms Num is the port number if not given the number is assigned sequentially The port is equivalent to a voltage source in series with a resistor The DC voltage of the source is supplied using vdc it specifies the DC voltage across the port when it is terminated in its reference impedance in other words vdc equals half the DC voltage supplied by the internal voltage source In an AC analysis mag and phase give the power delivered by the port when terminated with the reference resistance Similarly mag1 phasel mag2 give the power delivered at each of the harmonics in a harmonic analysis and mag1f phaself mag2f and phase2f give the power at each of the fundamentals in a spectral analysis Magnitudes are all given in dBm and phase is give in degrees DC voltage is modifiable ELEMENT DEFINITION Name n n port lt parameter value gt ELEMENT PARAMETERS r Reference resistance num Port number vdc DC vo
181. ng port of the CCVS since there can be more than one input port there must be as much of these parameters as the input ports or sensing devices This applies only if the CCVS is not modelled by an external macromodel In this case gains are specified by the macromodel rm Real transresistance This parameter is equivalent to lt gain1 gt sensedev String the name of the instance of the sensing device The current through this device is used to control the ccvs If the instance of the CCVS belongs to a subcircuit the sensing device is searched locally This means that is a local and global devices with the same name exist the local one is chosen There can be more than one of this parameter according to the number of input ports probe String the name of the instance of the sensing device The current through this device is used to control the ccvs If the instance of the CCVS belongs to a subcircuit the sensing device is searched locally This means that is a local and global devices with the same name exist the local one is chosen There can be more than one of this parameter according to the number of input ports gbsensedev String the global name of the instance of the sensing device The current through this device is used to control the ccvs Only global devices are considered devices at the top level of the netlist i e those not belonging to subcircuits are referred to as global If a global and a local device exist the
182. niently from a file The format of data is one point of the characteristic per row each point composed of two real numbers separated by a space the first represents voltage and the second the corresponding current quiet Boolean false it is a boolean parameter that when turned on suppreses the message about the file system path from which the file containing the pwl characteristic of the resistor is loaded The simulator searches and opens files by sequentially looking in the directory list specified through the environment variable lt PAN_PATH gt It opens the first file found in the directory list This can generate errors if files with the same name are contained in different directories For this reason a warning is given each time a file is found and opened tc1 Coefficient of the linear term in the temperature dependence expression of resistance Temperature affects only linear resistors tc2 Coefficient of the quadratic term in the temperature dependence expression of resistance Temperature affects only linear resistors tnom Nominal temperature at which resistance was determined rsh Real 0 0 sheet resistance narrow Real 0 0 narrowing of resistance etch Real 0 width narrowing due to etch per side short Real 0 0 shortening of resistance etchl Real 0 length narrowing due to etch per side w Real 1 0e 5 resistor with 1 Real 1 0e 5 resistor length af Real 1 0 flicker noise exponent kf Real 0 0
183. nitial guess The conventional solver adds the working frequency of the autonmous circuit among the problem unknowns and uses the Newton iterative method to solver the enlarge harmonic balance problem This method can not be employed by itself since it needs an initial guess This guess can be obtained for example by a previous harmonic analysis that employs a probe but that does not solve it or by time domain analyses such as trasient envelope shooting freltol Real 1 0e 6 the relative error tolerance in the computation of the working frequency of autonomous circuits vprobe1 List the list containing the pair of nodes at which the voltage probe on the first fundamental is connected If the second node of the list is omitted it is considered connected to ground The format is lt vprobel pos neg gt vprobeimagi Real 0 0 the magnitude of the voltage probe acting of the first funda mental vprobeiphi Real 0 0 the phase of the voltage probe acting on the first fundamental vprobeir1 Real 0 0 the value of the resistor connected in series to the voltage probe acting on the first fundamental vprobeimax Real the maximum value of the magnitude of the voltage probe employed during the sweep to find the steady state solution vprobe2 List the list containing the pair of nodes at which the voltage probe on the second fundamental is connected If the second node of the list is omitted it is cons
184. ns that leads to the steady state working condition of the circuit The circuit node branch that yields tha maximum relative error is displayed The algorithm computes also the eigenfunctions related to the variational model of the circuit at steady state The variational model is characterised by the fundamental matric M that admits the v T left eigenvectors and the u T right eigenvectors The v t and u t eigenfunctions along one working period T of the circuit are determined starting from the value of eigenvectors at the beginning of the period With respect to the state formulation these eigenfunctions are determined for each equation of the MNA formulation referring to nodes and currents of current controlled elements If the circuit is autonomous the eigenvector and thus the derived eigenfunction associated to the Floquet multiplier equal to 1 are automatically identified In general there can be more than one Floquet multiplier very close to 1 think of high Q oscillators that found as the representative of the Floquet multiplier theoretically equal to 1 is marked by the symbol in the multiplier printout The number of the Floquet multipliers differs from the number of state variables This happens since the modified nodal analysis formulation is adopted in the simulator Floquet multipliers refer to node voltages of voltage controlled elemets and branch currents of current controlled elements capacitors and
185. nt Count vl 0 vh VDD Voltage controlled switch sw implements also resistors R2 and R3 of the netlist in fact model SW vswitch ron 5 roff 100M von 0 6 VDD voff 0 4 VDD sets the values of roff and ron to those chosen respectively for Ro and R3 The AMS specific parts of this netlist are two the first is the single line verilog_include comb v that instructs the simulator to read also a Verilog file called in this case comb v more than one file can be read if needed This file shown below in Netlist 3 1 2 contains the description of the digital part of the circuit The second AMS specific part is used to connect the analog part to the digital one A2d1 x gnd a2d dignet cnt X vl 0 vh 0 D2al y gnd d2a dignet cnt Y vl 0 vh VDD D2a2 z gnd d2a dignet cnt Count vl 0 vh VDD These three line define three pseudo components that convert analog signals to digital or viceversa respectively a2d and d2a Node names are given in standard PAN netlist syntax while the corresponding digital net by using the dignet keyword Note that the format for digital net is module NET Voltage levels 3 1 A simple Analog Digital oscillator 37 must also be stated representing a threshold level for the a2d components and mapping levels for the d2a ones The digital part of the circuit is quite simple and written is standard Verilog language Netlist 3 1 2 The PERO oscillator digital part File
186. of the instance are compared to those of the models in the list When the width and lenght of the instance fall inbetween the min max values specified in one model of the list it is bound to the instance To summarise clusters are a way to choose models by using width and length of the given instance as a selection criterion bounderror Integer 2 the severity of the error given when lt I gt and or lt w gt are out of bounds lt Imin gt lt Imax gt or lt wmin gt lt wmax gt Possible choises are 0 none error is not given 1 warning simulation continues 2 error simulation is aborted Note that if this parameter is specified it overwrites the value of the corresponding parameter given among the lt options gt of the simulator type String the mosfet type Available keywords are lt n gt and lt p gt vbox Real 1 0e9 tox oxide breakdown voltage GaAs MESFET The GaAs MESFET model was derived from the model by H Statz et al at Raytheon GaAs MESFET instances require the use of a model statement There have been some convergence problems with this model in the harmonic and spectral analyses that result from Cgs going to zero beyond pinchoff The problem results when the gate is driven from an inductive source such as in traveling wave amplifiers and there is no other capacitance at the gate To prevent this problem it is suggested that Cgd not be set to zero and that side wall capacitance be added to the gate source an
187. olver is faster than the direct one but it can be less accurate The direct one is not suited for large circuits since it can take too much cpu time Note that the iterative solver is unable to compute all the poles of a large circuit The maximum number of poles in N 2 where N is the number of poles characterising the circuit The computed poles are ordered from that with the largest real part towards that with the smaller one maxm Real 1 0e12 poles or zeros are not considered when their module is greater than this parameter Recall that modified nodal analysis formulation introduces poles and zeros at infinity for circuit nodes of the algebraic portion of the DAE maxp Integer the maximum number of poles with largest real part that are computed by the iterative solver This parameter has meaning only when the iterative solver is employed Periodic Steady State Analysis The analysis computes the periodic steady state response of autonomous and nonautonomous circuits in the frequency domain by applying the Harmonic Balance Method Mem Waveforms made available to control blocks 138 Chapter 4 Reference 1 vector of employed fundamental and harmonics can be accessed by through the freq variable 2 spectrum of each node voltage and branch equation can be eccessed by postnoming the node branch equation name to the analysis name ANALYSIS DEFINITION Name harmonic lt parameter value gt ANALYSIS PARAMETERS annotate
188. omatically adjusts to read this format Also most simulator directives can be given in spice format but this feature is not completely supported and we suggest that native PAN syntax is used instead Also note that the order of insertion of each single directive in the final netlist is not simple to control and in some cases this may lead to errors e g when one simulation depends on the results of a previous one if a strict ordering of analyses is required we suggest an external file containing simulation directives is used or at least to review and eventual correction of the final netlist file Note that some quirk of the backend will put two identical INCLUDE lines in the output netlist you must delete or comment one of them in order to use the netlist 2 5 2 5 Steady state time domain analysis shooting 29 Netlist 2 4 1 The Bubba oscillator File bubba2 pan gnetlist g spice sdb o bubba2 pan 0 include_mode bubba2 sch Ae ee KK KK K K K K K FK K FK K K 2K K FK K K KK KK FKK KFK FK FK FK OK K FK K FK K K KK KK KK OK OK K K KE K 2K K K K Spice file generated by gnetlist spice sdb version 4 28 2007 by SDB provides advanced spice netlisting capability Documentation at http www brorson com gEDA SPICE KKK K 2K k 2 K K K K KK K K OK Kk K OK K 2 K 2K OK K K K K K OK K K K K K K K K K OK FK k K OK K K K K K OK K k kK k k INCLUDE TLV247xA sub x Begin SPICE netlist of main d
189. onal and defines the absolute tolerance when the norm of the residue is below this value it is assumed that convergence has been reached The returned values are the x vector that stores the found solution the real value err that stores the norm of the residue and the integer iters that stores the total number of performed iterations of the Newton algorithm norm x ones n m create a matrix of n rows and m column with all entries equal to 1 If the second argument is omitted a column vector of length n is created x pink n slope fsample generate colored noise with spectral density equal to f ope The n parameter tells how nany sample are needed The fsample parameter defines the frequency at which noise samples of colored noise are generated and thus the bandwidth of the generated noise x pinv y abs computes the pseudo inverse of the y matrix The abs argument defines the absolute tolerance below which the singular value value of y are not considered If abs is not given the singular considered are those greater than s1 gMachineResolution max M N where s1 is the largest singular value M is the number of rows of y and N is the number of columns f x mtpsd x bw nfft fs compute multitaper spectrum estimates of time sequence stored in the x vector bw is time bandwidth product its default value is bw 1 5 nfft defines the frequ
190. ond the scope of this simple walk through and is discussed in detail in many papers and more in general in several books see e g BGS09 BGS10 BMS05 Bra 10 2 5 Steady state time domain analysis shooting 33 Far94 Kuz95 Results of the tran and envelope analysis clearly showing the very long 0 3 Tran emitter Env emitter Tran base 2 Env base 4 F u 8 gt TT 8 F 10f L i 1 ri 1 2 L i i 0 1 2 3 4 0 0417 3 5417 7 0417 10 5417 time 4 0 time x1 0 Figure 2 14 The first short tran analysis for the Butler oscillator right panel and the long envelope analysis that follows left panel oscillator power up sequence are depicted in Figure 2 14 the initial tran analysis in the left panel the long envelope analysis in the right panel Details of this last analysis are shown in Figure 2 15 The envelope analysis arrives very close to the limit cycle i e steady state of the oscillator The final shooting analysis computes one cycle of the steady state waveform Results are shown in Figure 2 16 34 Chapter 2 A simple PAN walk through Env emitter Env emitter Env base Env base
191. onsidered linearly dependent on I The second mode to model a nonlinear current controlled inductor is through the polynomial 4 2 4 4 2 Basic one port device 45 Phi c0l c1 02P 031 c4P c51 where c0 cl c2 c3 c4 and c5 constant coefficients are specified in the model card if omitted default value of these parameters is 0 The third mode is similar to the second one but the nonlinear inductor is flux controlled I cOPhi c1Phi2 c2Phi gt c3Phi c4Phi gt c5Phi ELEMENT DEFINITION Name nl n2 ModelName lt parameter value gt ELEMENT PARAMETERS 1 Inductance It is considered only in a linear inductor icon Initial inductor current ic Initial inductor current fi String the algebraic function that describes the f v flux current characteristic of the inductor MODEL DEFINITION model ModelName inductor lt parameter value gt MODEL PARAMETERS type Inductor type 1 linear inductor 2 nonlinear saturable inductor 3 nonlinear current controlled inductor 4 nonlinear flux controlled inductor bmax Maximum induction magnetic field tesla at saturation It sets the value of the magnetic field when H N I Lis equal to Hmax where N is the number of turns I is inductor current and L is core length It is meaningful only for a saturable inductor corearea Magnetic core area meter meter It is the area of the cross section of the core and is meaningful only for a nonlinear saturable in
192. onstraints due to ic conditions It is not ensured that the specified values are met at convergence The list has the format lt ic node value branch value gt These initial conditions are implemented as Norton equivalents connected to the forced equations where the parallel resistor has a conductance equal to the product of the lt icforce gt parameter by the diagonal entry of the MNA matrix corresponding to the forced equation The magnitude of the independent current sources are so that to initially force the given initial conditions The Norton equivalents are swept out during the dc analysis icforce Real 1 0 the initial value of the coefficient employed to force the initial conditions for the Newton algorithm It has meaning effect only if the lt ic gt parameter is specified damping Boolean false apply a version of the damped Newton method srcstepping Boolean true turn on off the source stepping continuation algorithm gminstepping Boolean true turn on off the gmin stepping continuation algorithm ggroundstepping Boolean true turn on off the gground stepping continuation algorithm pseudotransient Boolean true turn on off the speudo transient continuation algorithm initgground Real this parameter has effect only when the pseudo transient continuation method is started The initial value of the parasitic conductance used to emulate the pseudo transient is set to the value specified by this parameter instead of
193. onver gence ereltol Real 1 0e 4 relative tolerance employed by the shooting analysis to check conver gence iabstol Real 1 0e 12 absolute current tolerance employed by the Newton algorithm at each time point of the time domain analysis ireltol Real 1 0e 6 relative current tolerance employed by the Newton algorithm at each time point of the time domain analysis vabstol Real 1 0e 6 absolute voltage tolerance employed by the Newton algorithm at each time point of the time domain analysis 4 8 Advanced time domain analyses 129 vreltol Relative voltage tolerance employed by the Newton algorithm at each time point of the time domain analysis trabstol Real 1 0e 12 absolute convergence tolerance at each time point of the transient analyses performed during the execution of the shooting analysis trreltol Real 1 0e 3 relative convergence tolerance at each time point of the transient analyses performed during the execution of the shooting analysis ntoltype Integer 2 option specifies the tolerance check adopted at each time point of the analysis to control if the Newton method has reached convergence Checks are done both on the variation of variables computed by the Newton method and on the entries of the right hand side vector Define as dX n the variation of the n electrical variable as reltol and abstol the relative and absolute maximum allowed variations respectively The n th vari
194. oreign simulators ignoreundefmodpars Boolean 0 undefined parameters of models are ignored This may be usefull to allow the reading of netlists of foreign simulators ignoreundefsubcktparams Boolean false parameters of subckt instances that are not de fined in the corresponding subckt body are ignored This may be usefull to allow the reading of netlists of oreign simulators ignoreundefparams Boolean false parameters that are not defined assigned before their usage are ignored or more precisely return a 0 value ignoreunknownoptions Boolean false unkown options are ignored bounderr Integer the severity of the error given when lt l gt and or lt w gt parameters of mosfets are out of bounds In general these bounds are defined by the lt Imin gt lt Imax gt and lt wmin gt lt wmax gt parameters Possible choises of this parameter are 0 none error is not given 1 warning simulation continues 2 error simulation is aborted Note that if this parameter is not specified bounds are checked according to the corre sponding action defined by the simular parameter of each mosfet model renamedupinst Boolean false rename duplicated instances that is those instances with names colliding with previously defined ones 4 7 2 4 7 3 4 7 Simulator control pseudo analyses 117 cc String cc O3 fPIC shared bhvcmp Boolean true the algebrain expressions of behavioral elements are compiled in a shared libra
195. ors In IEEE Transactions on Circuits and Systems I Fundamental Theory and Applications 52 4 Apr 2005 pages 681 694 cited on page 32 A Brambilla and G Storti Gajani Frequency Warping in Time Domain Circuit Simulation In IEEE Transactions on Circuits and Systems I Fundamental Theory and Applications 50 7 July 2003 pages 904 913 cited on pages 26 27 A Brambilla et al Improved Small Signal Analysis for Circuits Working in Periodic Steady State In Circuits and Systems I Regular Papers IEEE Transactions on 57 2 Feb 2010 pages 427 437 ISSN 1549 8328 DOI 10 1109 TCSI 2009 2019403 cited on page 32 Chung Wen Ho Albert E Ruehli and Pierce A Brennan The modified nodal approach to network analysis In Circuits and Systems IEEE Transactions on 22 6 June 1975 pages 504 509 ISSN 0098 4094 DOI 10 1109 TCS 1975 1084079 cited on page 12 Technical Reports MPO1 Ron Mancini and Richard Palmer Sine Wave Oscillator Technical report Texas Instruments Mar 2001 cited on page 27 i tin f i E A Set tied My if i 05 fle Se ay ii rt 1 ba l 7 ni Tem Ai 4 4 Pt wal nies ppe th o i ate i LT ee E teh 2h i Ses al oe te D Downloading Installing 7 37 envelOpe cece eee eee eee eens 25 Frequency Warping example 24 Restart option tran 00 23 shooting 06 eee eee eee eee ee 27
196. otal energy stored into the corresponding capacitor inductor The integration time step is chosen to keep variation with respect to total stored energy below the user defined threshold In general this method gives less accurate results with respect to the previous one but the analysis runs faster Integer 1 skip time point advance in time in case of unrecoverable convergence failures of the Newton algorithm It helps to continue and successfully terminate the simulation even though accuracy can be locally compromised Possible choises of this parameter are 0 no skipping is applyied in case there is no convergence of the Newton algorithms 1 skipping is applyed but only if there is convergence of the Newton algorithm in the new time point determined after the forward time skipping 2 forward time skipping is applyed anyway even though there is no convergence of the Newton algorithm in the new time point Boolean false if true parasitic capacitors of suitable values are inserted between each circuit node and ground when the simulator is in trouble to compute the solution This can aid integration algorithms maxdeltav Real the maximum allowed variation of node voltage allowed during the jump iterations of the Newton algorithm This should aid convergence and avoid possible overflow Boolean false activate finding of discontinuities in the vector field Saltation matri ces are exploited to construct a correct fudamental mat
197. ote how diode instance d1 is defined by the model template as is while diode d2 is based on model MYDIODE but has the additional imax parameter defined and parameter bv redefined instance parameters override those defined in the model Definition of new components will be discussed in a later section Global Options Global simulation options such as temperature convergence aid parameter values gmin gmax gground topology checker options global verbosity level and many more are defined with line name options optioni valuel option2 value2 where name is a optional unique identifier and as many space separated options as desired can be given after the options keyword Output format In general PAN will output to terminal and to a log file a considerable amount of messages These include error messages information on the simulator current activity and in some cases some information about simulation results Terminal output can be limited to error messages with option q i e prompt gt pan q netlist pan will result in a log file netlist log and terminal output containing only few notices con cerning simulator status and eventual error messages Simulation results for each analysis in netlist pan is stored in a directory named netlist raw This directory contains an ASCII 2 2 2 2 1 12 Chapter 2 A simple PAN walk through index file and one or more output file for each analysis each na
198. pecifies the net e of instance U1 that in turn is instantiated in the Mod root module bits list the list is composed of two items i e it has the form bits b1 b0 where b0 specifies the least significant bit of the digital net and b1 is the most significant one In the convertion only these bits are considered trtime Real the minimum allowed time window into which a variation of the analog output generated by the d2a conveter must happen vl Real the voltage value that defines the low logic state vh Real the voltage value that defines the high logic state vx Real 0 5 vh vl the voltage value that defines the x logic state This value must be less than vh and greater than v1 rout Real 1 0 the output port of the d2a is modeled as a driven current source with a resistor connected in parallel This parameter sets the value of the resistor Note that when the strength of the digital signal driving the d2a becomes Z the conductance of this resistor is automatically set equal to gmin rz Real 1 le 12 the output resistance of the d2a when the digital is in the z high impedance state If this parameter is not given that default value equals the reciprocal of the lt gground gt conductance analog to digital converter This element implements a pseudo analog to digital converter that links the electrical circuit to the digital circuit ELEMENT DEFINITI
199. pendence of alphal lbetaO Length dependence of beta0 lagidl Length dependence of agidl lbgid1 Length dependence of bgidl lcgidl1 Length dependence of cgidl legidl Length dependence of egidl laigc Length dependence of aigc lbige Length dependence of bigc lcige Length dependence of cigc 86 Chapter 4 Reference laigsd Length dependence of aigsd lbigsd Length dependence of bigsd lcigsd Length dependence of cigsd laigbacc Length dependence of aigbacc lbigbacc Length dependence of bigbacc lcigbacc Length dependence of cigbacc laigbinv Length dependence of aigbinv lbigbinv Length dependence of bigbinv lcigbinv Length dependence of cigbinv lnigc Length dependence of nigc lnigbinv Length dependence of nigbinv lnigbacc Length dependence of nigbacc 1Intox Length dependence of ntox leigbinv Length dependence for eigbinv lpigcd Length dependence for pigcd lpoxedge Length dependence for poxedge lvfbcv Length dependence of vfbcv lvfb Length dependence of vfb lacde Length dependence of acde lmoin Length dependence of moin lnoff Length dependence of noff lvoffcv Length dependence of voffcv lxrcrgi Length dependence of xrerg1 lxrcrg2 Length dependence of xrerg2 llambda Length dependence of lambda lvt1 Length dependence of vtl 1xn Length dependence of xn leu Length dependence of eu lvfbsdoff Length dependence of vfbsdoff ltvfbsdoff Length dependence of tvfbsdoff ltvoff Length dependence of tvoff wcdsc Width dependence of cdsc wcdscb Width
200. pendence of ub wub1 Width dependence of ub1 wuc Width dependence of uc wuci Width dependence of ucl wud Width dependence of ud wudi Width dependence of ud1 wup Width dependence of up wlp Width dependence of Ip wud Width dependence of u0 wute Width dependence of ute wvoff Width dependence of voff wminv Width dependence of minv wdelta Width dependence of delta wrdsw Width dependence of rdsw wrsw Width dependence of rsw wrdw Width dependence of rdw wprwg Width dependence of prwg wprwb Width dependence of prwb wprt Width dependence of prt weta0 Width dependence of eta0 88 Chapter 4 Reference wetab Width dependence of etab wpclm Width dependence of pclm wpdiblc1 Width dependence of pdiblc1 wpdiblc2 Width dependence of pdiblc2 wpdiblcb Width dependence of pdiblcb wfprout Width dependence of pdiblcb wpdits Width dependence of pdits wpditsd Width dependence of pditsd wpscbei1 Width dependence of pscbel wpscbe2 Width dependence of pscbe2 wpvag Width dependence of pvag wwr Width dependence of wr wdwg Width dependence of dwg wdwb Width dependence of dwb wbO Width dependence of bO wb1 Width dependence of b1 wcgsl Width dependence of cgsl wcgdl Width dependence of cgdl wckappas Width dependence of ckappas wckappad Width dependence of ckappad wef Width dependence of cf wclc Width dependence of clc wcle Width dependence of cle walphaO Width dependence of alpha0 walphai Width dependence of alphal wbeta0 Width depend
201. pendent current sources are so that the product of their currents by the parallel resistors is equal to the given initial conditions icforce Real 1 0 the initial value of the coefficient employed to force the initial conditions for the Newton algorithm It has meaning effect only if the lt ic gt parameter is specified fft Boolean false turns FFT analysis on off fftfund Real the frequency of the fundamental A fraction of the reciprocal of this value time period is employed to sample waveforms For more details see the lt fftperiod gt parameter fftperiod The period along which waveforms are sampled and FFT is performed The 114 Chapter 4 Reference sampling interval should begin lt fftperiod gt before the lt tstop gt value If the lt fftestper gt parameter is set true the lt fftperiod gt value is automatically adjusted within a maximum value of 40 fftestper Boolean false estimation of the period of waveforms along which the FFT transformation is applied If this parameter is true it is not mandatory to specify an accurate value of lt fftperiod gt since it is automatically adjusted within a maximum relative value of 40 fftharms Real 4 the number of harmonics computed by FFT fftsamples The number of sampling points in the FFT analysis This parameters is effective only if it specifies a number of samples greater than 4 lt ffthars gt 1 which represents the minimum allowed number of employ
202. possible choices are 0x0 noiseless JFET Ox1 thermal noise 0x2 shot noise 0x4 flicker noise The value of this parameter can be the result of the bitwise or of types of noise sources MODEL DEFINITION model ModelName jfet lt parameter value gt MODEL PARAMETERS nfet N type JFET njf N type JFET pfet P type JFET pjf P type JFET vto Pinch off voltage beta Transconductance parameter lambda Channel length modulation parameter rd Drain resistance rs Source resistance is Gate saturation current js Gate saturation current 4 4 4 4 4 semiconductor device 61 cgs Zero bias gate source junction capacitance cgd Zero bias gate drain junction capacitance pb Gate junction potential fc Forward bias depletion capacitance threshold tnom Parameter measurement temperature kf Flicker noise coefficient af Flicker noise exponent imax Explosion current Philips M903 mosfet transistor The m903 mosfet transistor developes by Philips ELEMENT DEFINITION Name drain gate source bulk ModelName lt parameter value gt ELEMENT PARAMETERS w W 1L mult Integer 1 number of identical mosfets connected in parallel m Integer 1 number of identical mosfets connected in parallel as Real source area ad Real drain area ps Real source perimeter pd Real drain perimeter MODEL DEFINITION model ModelName m903 lt parameter value gt MODEL PARAMETERS nmos Nmos pmos Pmos ler LER wer WER
203. r vtsswgd Drain gate edge sidewall trap assisted voltage dependent parameter gbmin Minimum body conductance rbdb Resistance between bNode and dbNode rbpb Resistance between bNodePrime and bNode rbsb Resistance between bNode and sbNode rbps Resistance between bNodePrime and sbNode rbpd Resistance between bNodePrime and bNode rbpsO Body resistance RBPS scaling rbpsl Body resistance RBPS L scaling rbpsw Body resistance RBPS W scaling rbpsnf Body resistance RBPS NF scaling rbpdO Body resistance RBPD scaling rbpdl Body resistance RBPD L scaling rbpdw Body resistance RBPD W scaling rbpdnf Body resistance RBPD NF scaling rbpbx0 Body resistance RBPBX scaling rbpbx1 Body resistance RBPBX L scaling rbpbxw Body resistance RBPBX W scaling rbpbxnf Body resistance RBPBX NF scaling rbpby0O Body resistance RBPBY scaling rbpbyl Body resistance RBPBY L scaling rbpbyw Body resistance RBPBY W scaling rbpbynf Body resistance RBPBY NF scaling rbsbx0 Body resistance RBSBX scaling rbsby0 Body resistance RBSBY scaling rbdbx0 Body resistance RBDBX scaling rbdby0 Body resistance RBDBY scaling rbsdbxl Body resistance RBSDBX L scaling rbsdbxw Body resistance RBSDBX W scaling rbsdbxnf Body resistance RBSDBX NF scaling rbsdbyl Body resistance RBSDBY L scaling rbsdbyw Body resistance RBSDBY W scaling rbsdbynf Body resistance RBSDBY NF scaling lcdsc Length dependence of cdsc lcdscb Length dependence of cdscb 84 Chapter 4 Reference
204. r is badly controlled and can lead to a failure of the Newton method Once more experience shows that the Newton algorithm by its intrinsic iterative nature is almost always able to dump the effects of the round off errors but still there can be circuit yielding to a very bad conditioned Jacobian matrix that can cause convergence problems skipdc Boolean false the initial DC analysis performed to determine the initial conditions is skipped This can help when dealing with critical circuits that cause failures of the DC 112 Chapter 4 Reference analysis In this case voltages across capacitors and current through inductors are assumed null uic Integer 0 use given initial conditions of inductors and capacitors Initial conditions are considered in two different ways according to the value of this paramenter in particular 1 it is assumed that all the capacitors and inductors have suitable initial conditions This means that if they are not specified by the user they are assumed equal to zero Note that this initial condition determination fails when there are loops of voltages source and capacitors even though consistent initial conditions are given since the loop current is undertermined The same applies to inductor cut sets 2 itis assumed that all the capacitors and inductors whose initial conditions have not been specified are open circuit and short circuit respectively restart Boolean true restart the analysis that is do no
205. r of types of noise sources MODEL DEFINITION model ModelName poly lt parameter value gt MODEL PARAMETERS af Real 1 0 flicker noise exponent kf Real 0 0 flicker noise coefficient semiconductor device Junction Diode The diode is tied between two pins anode and cathode The other three optional pins ther mal didy and didT are internally connected to independent current sources The thermal pin refers to the heat flow from the diode and is active only during an eletro thermal analysis It must be connected to a node of the thermal network Heat flow is equal to the electrical power through the electrical port and is modeled by a current source tied between ground and the thermal node the related current exits from the thermal node The didT pin refers to the derivative of the diode branch current with respect to diode branch voltage Finally the didT pin refers to the derivative of the diode branch current with respect to juntion temperature These optional pins may be made inactive if connected to ground ELEMENT DEFINITION 4 4 semiconductor device 57 Name anode cathode thermal didv didT ModelName lt parameter value gt ELEMENT PARAMETERS area Junction area in square meters perimeter Junction perimeter in meters peri Junction perimeter in meters perim Junction perimeter in meters pj Junction perimeter in meters region DC operating regio
206. rameter has not been specified the period specified or computed by the previous analyses is adopted printnodes Print external internal node map printsh Boolean no when true the solution obtained at convergence of the shooting method is printed savelist It defines a list on nodes or branch currents that will be saved into the output file during simulation Other computed electrical variables are not saved thermalnet Boolean false use the thermal network checkstrange Boolean 1 when true at each time point of the transient analysis it is checked if there are some strange behaviours of devices For example it is checked if bulk drain and bulk source parasitic diodes of mosfets are forward biased This check slows down a bit simulation and can generate a large number of warnings autonomous Boolean no it tells if the circuit is autonomous In this case the working period of the circuit is automatically estimated vprobe List the list containing the pair of nodes at which the voltage probe is connected If the second node of the list is omitted it is considered connected to ground The format is vprobe pos neg probeharm Real 1 the index of the harmonic at which the probe drives the circuit vprobemax Real the maximum value of the magnitude of the voltage probe employed during the sweep to find the steady state solution eabstol Real 1 0e 6 absolute tolerance employed by the shooting analysis to check c
207. rcuit Since we are looking for a periodic steady state in the non autonomous case the circuit is periodically forced with known period T on the other hand if the circuit is autonomous period T must also be determined The shooting algorithm requires computing the sensitivity of the final solution with respect to the initial conditions This is quite straightforward if all equations are continuous but requires some work otherwise The Butler overtone oscillator The oscillator used in this section known as the Butler overtone oscillator is taken from Rhe95 and is shown in Fig 2 13 It is a crystal oscillator the crystal is modelled as a simple third order RLC subcircuit that in the PAN netlist shown below in 2 5 1 is named xtal The bjt model for the 2N5179 device has been taken from the National Semiconductors site Netlist 2 5 1 The Butler oscillator File butler pan ground electrical 0 options gground 1 0e 40 topcheck 2 parameters FREQ 120M Period 1 FREQ 3The method is called shooting for this reason it is analogous to the adjustments that a gunman makes to his initial cannon tilt and yaw i e the initial guess in order to reach the target In our case the gunman is actually firing a boomerang expecting it to come back in the same point where it was thrown 2 5 Steady state time domain analysis shooting 3 Analyses De de print yes Tran tran tstop 500 Period uic yes method 2 order 6
208. rcuits to oscillate or even to become unstable skip Integer 1 skip time point advance in time in case of unrecoverable convergence failures of the Newton algorithm It helps to continue and successfully terminate the simulation even though accuracy can be locally compromised Possible choises of this parameter are 0 no skipping is applyied in case there is no convergence of the Newton algorithms 1 skipping is applyed but only if there is convergence of the Newton algorithm in the new time point determined after the forward time skipping 2 forward time skipping is applyed anyway even though there is no convergence of the Newton algorithm in the new time point ic List specifies the initial value of node voltages and branch currents employed in the tran sient analysis These values override those automatically computed by the lt dc gt analysis as initial guess A dc analysis is initially done where the ic conditions force the Newton method It is not ensured that the specified values are met at convergence even though the result should be close to them The list has the format lt ic node value branch value gt For circuit nodes these initial conditions are implemented as Norton equivalents con nected to the forced node where the parallel resistor has a conductance equal to the product of the lt icforce gt parameter by the diagonal entry of the MNA matrix corresponding to the forced equation The magnitude of the inde
209. rease efficiency and robustness The DC analysis can look for possible multiple solutions The implemented method has been described in G Storti Gajani A Brambilla A Premoli Numerical Determination of Possible Multiple DC Solutions of Nonlinear Circuits IEEE Tranaction on CAS I Vol 55 No 4 May 2008 pp 1074 1083 ANALYSIS DEFINITION Name de lt parameter value gt ANALYSIS PARAMETERS annotate Integer 3 degree of annotation Values from 1 to 7 turn on display of generic information about the analysis run Values above 7 are bitwise and turn on specific printouts 0x08 at each iteration of the Newton method some information about the partial solution that maximally violates accuracy thresholds is displayed 0x10 at each iteration of the Newton method values of partial solutions are displayed 0x20 at each iteration of the Newton method information about current values of partial solution their increments and values of RHS is displayed 0x40 print MNA matrix RHS vector and the DeltaSolution vector at each iteration of the Newton method 0x80 dump MNA matrix in the analysis_name m file at each iteration of the Newton method 0x100 the histogram of the size of sub circuits loops and branches handled or identified by the circuit partitioner is displayed An example of the format displayed is Loops 2 3467 8 where 2 refers to the loop number is the histogram 3467 is the the number o
210. ressions PAN expressions can be formed using the standard four arithmetic operators plus a number of standard mathematical functions for a full list type 24 Chapter 2 A simple PAN walk through prompt gt pan s expressions The output is somewhat terse but should allow to solve most problems one may encounter Expressions are allowed in all fields where a number is expected The nonlinear transconductance device poly PAN also allows to define rational transconductances These are defined using device poly It is a nonlinear voltage controlled current source two port The first port i e the first two node names represent the current source while the second pair the sense nodes In our case since we want to realize a simple non linear one port the two ports are the same Coefficients for numerator and denominator of the rational function are given using parameters nO n1 n8andd0 di d8 respectively Not specifying any denominator coefficient is equivalent to d0 1 on the other hand not specifying any numerator coefficient results in anull polynomial This means that polynomials up to degree 8 can be defined In our case the netlist line ri n20 0 n20 O poly n0 0 ni 1 R n2 0 n3 ALPHA R d0 1 represents function i a Rv gt 1 Rv Voltage sources as current measure devices A null independent voltage source has also been introduced in series with the non linear element this forces MNA to explicitly add t
211. ription Pl Pw and Pp are not shown but can be recognized The names of P1 Pw and Pp are identical to that of PO but with a prefix of pl pw and plw respectively ELEMENT DEFINITION Name drain gate source ModelName lt parameter value gt ELEMENT PARAMETERS w Channel width 1 Channel length nrd Drain squares nrs Source squares m Multiplicity factor number of MOSFETs in parallel region Estimated operating region Possible values are off triode sat or subth noisetype Integer 5 types of generated noise possible choices are 0x0 noiseless resistor 0x1 thermal noise 0x4 flicker noise The value of this parameter can be the result of the bitwise or of types of noise sources rthO Thermal resistance cthO Thermal capacitance MODEL DEFINITION model ModelName psitft lt parameter value gt MODEL PARAMETERS type Transistor type Possible values are n or p vto Threshold voltage at zero body bias BIN lambda Real 0 048 channel length modulation parameter BIN tox Gate oxide thickness eta Subthreshold ideality factor BIN etai Alias to eta BIN asat Proportionality constant of Vsat BIN alphasat Alias to asat BIN delta Transition width parameter BIN mus Subthreshold mobility BIN mud High field mobility BIN 4 5 misc device 101 muo Alias to mu0 BIN mut Low field mobility parameter BIN mmu Low field mobility exponent BIN m Alias to mmu BIN vfb Flat ban
212. rix of circuits characterised by 130 Chapter 4 Reference discontinuities in the vector field Discontinuities in the vector field can be also intro duced by sharp varying functions modeling circuit elements such as for example the y tanh a x function with the a gain too large The solver automatically determines possible discontinuities due to numerical problems introduced by too sharp nonlinear func tions and automatically introduce saltation matrices The solver introduces also saltation matrices due to dscontinuities caused by commutations of pseudo analog to digital and digital to analog converters during simulations of mixed digital analog circuits and due to behavioural elements tagged as digital jumpabstol Real 1 0e 6 absolute tolerance employed to determine possible jumps in the vector field current through capacitors and voltage across inductors jumpreltol Real 1 0e 3 relative tolerance employed to determine possible jumps in the vector field current through capacitors and voltage across inductors floquet Integer 0 the Floquet multipliers related to the state transition matrix computed at convergence of the shooting method are evaluated Possible values of this parameter are 1 the Floquet multipliers are shown after convergence of the shooting algorithm 2 the Floquet multipliers are shown at each iteration of the shooting algorithm printeig Boolean no print the left and right eigenvectors
213. rough the statement get ShootName freq 3 device current waveforms the related time samples can be retrieved in the devtime vector 128 Chapter 4 Reference 4 Floquet periodic eigenfunctions that can be retrived through the statement get ShootName ul n1 where ul is the type of eigenfunction and n1 is the referring circuit node the related time samples are stored in the eigtime vector ANALYSIS DEFINITION Name shooting lt parameter value gt ANALYSIS PARAMETERS annotate Degree of annotation It is an integer number from 1 to 7 that turns on displaying of generic information about the analysis run Values above 7 are bitwise and turn on specific printouts In particular 0x08 print sensitivity matrix and RHS at various phases of the shooting analysis 0x10 at convergence of the shooting method the fundamental matrix is saved in a file named as the analysis 0x20 give a short report on actions performed to find saltation matrices maxiters Integer 100 maximum allowed iterations for the shooting algorithm restart Boolean yes restart the shooting analysis that is do not use previous results computed by a transient analysis an envelope analysis or a shooting analysis as initial guess to the current shooting analysis If the circuit is autonomous the period computed by one of the previous analyses is adopted as initial guess If the circuit is non autonomous and the lt period gt pa
214. rrent value in the inductor is set to 0 After simulation the command is pan vanderpol pan from now on we will specifiy the simulation command only when non standard we find the folder vanderpol raw containing the usual index file and files TranO and Trani containing simulation results Plotting the inductor current versus the capacitor voltage yields the trajectory shown in Fig 2 3 The time domain analysis tran 25 KN oO A 4 3 2 1 al Figure 2 7 Inductor current vs capacitor voltage for the Van der Pol oscillator Netlist 2 3 1 Restarting from previous results As desired simulation starts from initial conditions set at V 3V and Iz Q It is sometimes useful to start a time domain simulation from the last point of a previous simulation This can be done using option restart whose default value is yes as in the second analysis line Trani tran tstop 120 restart no setting restart no instructs the simulator to initiate this time domain simulation using as initial point the last point computed in the immediately previous tran shooting or envelope simulation Note that when restarting from a previous simulation option tstop must take into N wo A 4 3 2 14 0 1 2 3 4 V C Figure 2 8 Inductor current vs capacitor voltage for the Van der Pol oscillator Netlist 2 3 1 account the stop time of the previous simulation in our case setting tstop 120 means that 2 3 2 26 Chapter
215. rresponding coupling coefficient The first inductance of this sequence is taken as the reference inductor the other ones are those to which the reference inductor is coupled to Note that coupling is mutual ELEMENT DEFINITION Name coupledind lt parameter value gt ELEMENT PARAMETERS indi ind2 This parameter specifies the instance of another inductor to which there is an inductive coupling There must be two ore more of these parameters since coupling can take place only between two or more inductors The first specified inductor is taken as reference Note that the other specified inductor is automatically mutually coupled to the reference one ki k2 This parameter specifies the coupling coefficient between two inductors Its value must be in the range 1 1 Polynomial Transconductor The poly element is a 4 terminal nonlinear resistive element whose output current is given by i n0 nlxv n2 v n8x v d0 d1 v d2 v d8 v where v is the voltage between the sense nodes in volt and i is the output current in ampere Positive current exits the source node and enters the sink node Noise is considered only if the element behaves like a nonlinear polinomial resistor that is the sink source nodes coincide with the sense sense ones 4 4 4 4 1 56 Chapter 4 Reference ELEMENT DEFINITION Name sink source sense sense ModelName lt parameter value gt ELEMENT PARAMETERS n0 Re
216. ry This shared library is loaded and executed and executed by the simulator temperature Real 27 0 circuit working temperature in Celsius degrees temp Real 27 0 circuit working temperature in Celsius degrees subckthier Boolean false print subcircuit hierarchy outintnodes Boolean false when set voltages of internal nodes are saved into the output rawfile Internal nodes are added by devices such as mosfets diodes bjts and others to deal with internal elements or equations For example a diode is characterised at its extenal nodes those specified in the netlist by a series connection of a parasitic resistor and the diode itself The node connecting the parasitic resistor and the diode is considered as internal to the diode model and in general of no meaning to the user hidden to the user writeparams Boolean false write parameter values digits Integer 4 print precision width Integer screen width ascii Boolean false generate ascii rawfile resourceusage Integer print brief summary of resource usage timer Integer 1800 time interval after which the partial results of the currently running analysis are saved in a temporary file In case of failure or crash of the analysis the simulator can be restarted from this saved result Alter an instance parameter or simulation option This statement changes the value of a modifiable device instance parameter for any analyses that follow Only one param
217. s can be accessed through the command get EnvelopeName node 2 The indices vector that stores the indices of the first sample in each time window The length of this vector equals the number of time windows of the envelope analysis Indices can be accessed through get EnvelopeName indices ANALYSIS DEFINITION Name envelope lt parameter value gt ANALYSIS PARAMETERS annotate Degree of annotation It is an integer number from 1 to 7 that turns on the display of generic information about the analysis execution Values above 7 are bitwise and turn on specific printouts In particular 0x08 print sensitivity matrix and RHS at various phases of the analysis 0x10 write to an output file the sensitivity matrix at each successful iteration of the envelope algorithm checkstrange It is a boolean parameter default false When it is set true at each time point of the transient analysis it is checked if there are some strange behaviours of devices For example it is checked if bulk drain and bulk source parasitic diodes of mosfets are forward biased This check slows down a bit simulation and can generate a large number of warnings printnodes Print external internal node map autonomous It is a boolean flag default false that tells if the circuit is autonomous if set to true This causes the automatic estimation of the working period of the circuit statevars It is a list of names of device instances for exampl
218. s is a convergence aid for algorithms If it set to 0 its effects are turned off period Real the known working periods of the circuit in case of a nonautonomous circuit or an estimation of the working period in cae of autonomous circuits This estimation is refined by the envelope analysis till reaching the exact working period of the circuit fund Real the fundamental frequency of the circuit in case of nonautonomous circuits or an estimation of the working frequency in case of autonomous circuit This value is refined till finding the exact value by the envelope analysis restart If set to false results obtained by a previous envelope or transient analysis are uses as initial guess to the shooting method thermalnet Use the thermal network tinc Lengthening factor of the time step at convergence Default value is 1 1 tmax Maximum allowed time step lenght tmin Minimum allowed time step lenght savelist It defines a list on nodes or branch currents that will be saved into the output file during simulation Other computed electrical variables are not saved stop Time stop or ending time instant of the envelope analysis This parameter is ignored in shooting analysis tstop Time stop or ending time instant of the envelope analysis This parameter is ignored in shooting analysis uic Integer 0 use given initial conditions of inductors and capacitors Initial conditions are considered in two different ways according to the value of this
219. s of the lt spectral gt analyses It is composed of two items The former is the index of the harmonic of the first fundamantal and the latter is the index of the harmonic of the second fundamental For example freq2t 1 1 defines the frequency 1 f1 1 f2 in the lt spectral gt analysis p2mag Real 0 0 the magnutude of the second current probe in the lt spectral gt analysis p2ph Real 0 0 the phase of the second current probe in the lt spectral gt analysis p2g Real 0 0 the value of the conductance connected in parallel to the current probe acting on the second fundamental and its harmonics 4 5 4 4 5 5 4 5 misc device 97 Microstrip Line A microstrip line based on the equations of Hammerstad and Jensen The model contains a the thickness correction to the width and frequency dependent permittivity and characteristic impedance The dispersion equations are those of Kirschning and Jansen ELEMENT DEFINITION Name port port2 gnd msline lt parameter value gt ELEMENT PARAMETERS 1 Length in meters w Width in meters h Substrate height in meters t Conductor thickness in meters eps Substrate permittivity relative to a vacuum Linear N Port The characteristics of the N port versus frequency are determined by reading an S parameter data file S parameters maybe entered in one of the following formats real imag mag deg mag rad db deg db rad If the frequency scale factor is given it is multiplie
220. s the r v resistance voltage characteristic of the resistor 1 Real 0 0 resistor length w Real 1 0e 5 resistor width tc Real 0 coefficient of the linear term in the temperature dependence expression of resistance Temperature affects only linear resistors If a resistor model is associated to the instance and this parameter is specified only in the model the instance inherits the value given in the model tcl Real 0 coefficient of the linear term in the temperature dependence expression of resis tance Temperature affects only linear resistors If a resistor model is associated to the instance and this parameter is specified only in the model the instance inherits the value given in the model tc2 Real 0 coefficient of the quadratic term in the temperature dependence expression of resistance Temperature affects only linear resistors If a resistor model is associated to the instance and this parameter is specified only in the model the instance inherits the value given in the model noisetype Integer 5 types of generated noise possible choices are 0x0 noiseless resistor 0x1 thermal noise 0x4 flicker noise The value of this parameter can be the result of the bitwise or of types of noise sources MODEL DEFINITION model ModelName resistor lt parameter value gt MODEL PARAMETERS r Real 1 0 the value of resistance is applied to all resistors referring to this model the resistance value of each
221. s with the same name exist the local one is chosen There can be more than one of this parameter according to the number of input ports This parameter applies when the NPORT characteristic is described by algebraic functions gbsensedev String the global name of the instance of the sensing device The current through this device is used to control the NPORT Only global devices are considered devices at the top level of the netlist i e those not belonging to subcircuits are referred to as global If a global and a local device exist the global one is selected If the global one does not exist an error is displayed There can be more than one of this parameter according to the number of input ports This parameter applies when the NPORT characteristic is described by algebraic functions sensen String list the list composed of two nodes forming a pair across which voltage is sensed This voltage is used to control the NPORT Pairs are identified by a label formed by the n character followed by an integer numbering the pair Pairs are numbered from integer 1 in increasing order as they are defined in the NPORT card Intead of a list of two nodes a string specifing 1 node can be given In this case the pair is formed by the given node and ground MODEL DEFINITION model ModelName nport lt parameter value gt MODEL PARAMETERS damping Boolean true automatically tag some equations of macro models in order to turn on a
222. sient continuation method mindevres Real 1 0e 3 the minimum allowed value that the possible resistors connected in series to the internal device nodes can assume For example a diode has an internal resistor that models resistance of contancts and silicon that do not constitute the junction When a resistor internal to a device model has a value less than that specified by this parameter it is not considered and removed from the circuit pivrelthresh Real 0 001 this parameter determines what the pivot threshold will be It should be between zero and one If it is one then the pivoting method becomes complete pivoting which is very slow and tends to fill up the matrix If it is set close to zero the pivoting method becomes strict Markowitz with no threshold The pivot threshold is used to eliminate pivot candidates that would cause excessive element growth if they were used Element growth is the cause of roundoff error Element growth occurs even in well conditioned matrices Setting lt pivrelthresh gt large will reduce element growth and roundoff error but setting it too large will cause execution time to be excessive and will result in a large number of fill ins If this occurs accuracy can actually be degraded because of the large number of operations required on the matrix due to the large number of fill ins A good value seems to be 0 001 This value should be increased and the matrix resolved if growth is found to be excessive Changing
223. since this circuit is autonomous also its period must be determined and option autonomous yes must be explicitly specified The steady state behaviour of the circuit is shown in Fig 3 3 where the state variables are shown as a trajectory in state space on the z axis we have put in this case the binary counter value and Fig 3 4 where they are shown versus time 38 Chapter 3 AMS simulation with PAN 0 5 7 Wii Oo 0 50 100 150 200 250 300 Figure 3 2 The time evolution of the three state variables of the PERO oscillator Capacitor voltage is x in the top panel inductor current I1 1i in the centre panel and the analog signal corresponding to the digital two bit counter value remapped in the 0 to 1 range in the bottom one The pnoise analysis In the previous example based on the PERO oscillator after the shooting analysis a pnoise analysis is also performed The relevant line is Pnoise pnoise start 10u stop 100m dec 4000 onodes x sweeptype 0 This analysis requires a previous shooting analysis and computes the noise at a specified output in our case node x of circuits that admit a periodic solution 3 2 The pnoise analysis 39 Figure 3 3 Steady state trajectory of the PERO oscillator plotted in state space 0 2 0 2 0 5 Wii O 0 5 i i i i N 05H 300 305 310 315 320 325 ti
224. st of time instants at which a time domain analysis must be performed In case of non autonomous circuits the value of period is given by the user and does not vary during the envelope analysis In case of autonomous circuits the period varies and is automatically determined by the envelope analysis In both cases the time domain analyses are performed along time windows whose initial time instants are as close as possible to time instants specified in the list This is achieved by suitably choosing the envelope integration time step H eabstol Real 1 0e 6 absolute tolerance to check acceptability of the solution found by the envelope analysis at each working period of the circuit ereltol Real 1 0e 1 relative tolerance to check acceptability of the solution found by the envelope analysis at each working period of the circuit iabstol Absolute current tolerance employed by the Newton algorithm at each time point of the time domain analysis default 1 0e 12 ireltol Relative current tolerance employed by the Newton algorithm at each time point of the time domain analysis vabstol Absolute voltage tolerance employed by the Newton algorithm at each time point of the time domain analysis vreltol Relative voltage tolerance employed by the Newton algorithm at each time point of the time domain analysis trabstol Absolute convergence tolerance of the transient analysis employed by the envelope analysis trreltol Relative conv
225. t can be used by the transient analysis Noise analysis of a periodic circuit The pnoise analysis computes the noise at the specified output for example a node pair of circuits that admit a periodic solution These circuits can be both autonomous oscillators and non autonomous The pnoise analysis linearises the circuit along the limit cycle defined by the periodic large signal solution Non linear elements are transformed in time varying linear ones that thus exibit frequency translations The pnoise analysis must be preceeded by a shooting analysis that determines the limit cycle of the circuit The folding of noise sources is considered and thus the total noise at the specified output is composed of various contributions due to up conversion and down converions of noise The number of beats considered in folding noise can be specified through an analysis parameter An example of a pnoise analysis card is Pnoise pnoise start 50k stop 4M lin 100 onodes out The Pnoise analysis determines the total noise at the out node of the circuit a linear sweep of the frequency composed of 100 samples is started at a frequency offset from the fundamental equal to 50k The sweep ends at a frequency far 4M from the fundamental The value of the fundamental is automatically determined by the shooting analysis preceeding the pnoise one A pnoise analysis can be carried out also at specific frequency offsets given through a
226. t parameter for C V model cle Vdsat parameter for C V model dwc Delta W for C V model dlc Delta L for C V model alphaO substrate current model parameter alphai substrate current model parameter beta0O substrate current model parameter lcdsc Length dependence of cdsc lcdscb Length dependence of cdscb lcdscd Length dependence of cdscd lcit Length dependence of cit 1Infactor Length dependence of nfactor 1xj Length dependence of xj lvsat Length dependence of vsat lat Length dependence of at 1a0 Length dependence of a0 lags Length dependence of ags lai Length dependence of al 4 4 semiconductor device 71 la2 Length dependence of a2 lketa Length dependence of keta l1nsub Length dependence of nsub Inch Length dependence of nch lngate Length dependence of ngate lgammai Length dependence of gammal lgamma2 Length dependence of gamma2 lvbx Length dependence of vbx lvbm Length dependence of vbm 1lxt Length dependence of xt 1k1 Length dependence of k1 1kt1 Length dependence of kt1 1kt11 Length dependence of kt11 1kt2 Length dependence of kt2 1k2 Length dependence of k2 1k3 Length dependence of k3 1k3b Length dependence of k3b 1w0 Length dependence of w0 1nlx Length dependence of nlx ldvtO Length dependence of dvt0 ldvt1 Length dependence of dvt1 ldvt2 Length dependence of dvt2 ldvtOw Length dependence of dvtOw ldvtiw Length dependence of dvtlw ldvt2w Length dependence of dvt2w ldrout Length dependence of drout
227. t use results computed by a previ ous transient envelope or shooting analysis as initial condition for the current transient analysis save String the path of the file into which the solution is saved bz2 format is adopted The solution stored in this file can be later loaded by the simulator and used as initial condition from which the transient algorithm is started load String the path of the file from which a previously saved solution is read The file format can be both ascii and bz2 and it is automatically determined during loading This solution is used as initial condition to the transient algorithm checklc Boolean false find check and reduce trees of large conductances This can be an aid to convergence of the integration algorithm since large conductance reduction prevents growing of round off errors lclocal Boolean false use local algorithm for large conductance reduction tmin Real minimum allowed time step length The value of this parameter is automatically determined by the simulator if it is not specified by the user tmax Real or List maximum allowed time step length If a list is given it has the format t1 t2 ts1 The value of this parameter is automatically determined by the simulator if it is not specified by the user tinc Real 1 1 lengthening factor of the time step at convergence timepoints List the list containing the time instants that should be employed by the simula
228. tact ldif ACM Parameter Length of LDD Gate Source Drain 1d ACM Parameter Length of LDD under Gate paramchk Model parameter checking selector rd ACM Parameter Resistance of LDD drain side rs ACM Parameter Resistance of LDD source side rdc ACM Parameter Resistance contact drain side rsc ACM Parameter Resistance contact source side version String 3 2 2 model version Supported ones are 3 2 4 3 2 3 3 2 2 3 2 xl Length correction parameter xw Width correction parameter saref Reference distance between od edge to poly of one side sbref Reference distance between od edge to poly of one other side kvthO Reference distance between od edge to poly of one other side 1lkvthO Threshold shift parameter for stress effect stk2 k2 shift factor related to vthO change 1ku0 Length dependence of ku0 ku0 Mobility degradation enhancement coefficient for stress effect tku0 Temperature coefficient of ku0 kvsat saturation velocity degradation enhancement parameter for stress effect steta0O Eta0 shift factor related to vthO change tlev DC temperature selector tlevc DC temperature selector type String the mosfet type Available keywords are lt n gt and lt p gt Berkeley Short Channel IGFET Model Version 4 50 ELEMENT DEFINITION Name drain gate source bulk ModelName lt parameter value gt ELEMENT PARAMETERS 1 Length 4 4 semiconductor device 77 w Width m Integer 1 0 number of devices connecte
229. the pivot threshold does not improve performance on matrices where growth is low as is often the case with ill conditioned matrices pivabsthresh Real 0 0 the absolute magnitude a matrix element must have to be considered as a pivot candidate This number should be set significantly smaller than the smallest diagonal element that is expected to be placed in the matrix This element has a value equal to the lt gground gt parameter If there is no reasonable prediction for the lower bound 116 Chapter 4 Reference on these elements then lt pivabsthresh gt is set equal to the square of lt gground gt The lt pivabsthresh gt is used to reduce the possibility of choosing as a pivot an element that has suffered heavy cancellation and as a result mainly consists of roundoff errors pivcaching Boolean true when true the chaching mechanism to gather pivots during the LU factorization of the MNA matrix is adopted Pivot chaching augments the effeciency in LU factorizing the matrix In general the chaching mechanism does not worsen the conditioning of the MNA matrix but in extreme cases this can possibly happen In this case it is suggested to turn off the pivot chaching annotate Integer set annotation level of simulator Verbosity is expressed through an exadecimal number higher bit levels augment annotation bit enumeration starts from 0 the meaning of each bit is the following bit 1 turns on default simulator annotation bits 2
230. their output voltages These variations are checkes independently from their magnitudes each time there is a variation between two subsequent integration time instants this variation is forced to happen in less than lt trtime gt value In other words output voltage rise and falling times are ensured to be less than lt trtime gt This parameter helps the simulator to converge since when there is a transition of a lt digital gt CCVS the output voltage in theory must switch instantaneously From the simulator point of view this meas to shrink the integration time step till the minimum value allowed by the finite precision of the computer arithmetic unit Obviously this is time consuming since the length of the integratio time step governs the simulation speed To avoid this drawback if the lt trtime gt is specified it defined the minimum integration time step that is lt trtime gt below which the transition can be considered satisfing deri Real 0 0 the CCVS can act as a differentiator the branch voltage of the controlled port port 1 can be equal to the derivatives of the current through the controlling ports port 2 port3 multiplied by the lt der gt values Since there can be more than one controlling port this parameter can be specified for each controlling port Note that derivatives are considered as additive with respect to the input output gains These parameters apply only if the CCVS is not modelled by an external ma
231. tion current js Bulk juncction saturation current density tox Oxide thickness 1d Lateral diffusion uo Surface mobility fc Forward bias juction fit parm nsub Substrate doping tpg Gate type nss Surface state density delta Width effect on threshold vmax Maximum carrier drift velocity xj Junction depth nfs Fast surface state density eta Drain surce voltage effect on the threshold voltage xd Depletion layer width delta Width effect on threshold theta Mobility modulation alpha Alpha kappa Kappa tnom Parameter measurement temperature Celsius vt0O Threshold voltage vto Threshold voltage w Default width of the mos model rsh Sheet resistance rds Real infinite the resistance connected between drain and source terminal internal ones of the mosfet rd Drain ohmic resistance default 10 mOhm rs Source ohmic resistance default 10 mOhm Berkeley Short Channel IGFET Model Version 3 The bsim3 mosfet transistor model developed at the Berkeley University Clusters of this model are supported A cluster is a list of models when an element instance refers to a cluster of models the list is searched The lt wmin gt lt wmax gt lt Imin gt and lt lmax gt values speficied in each model of the cluster are compared to the current values of the lt w gt and lt l gt parameters of the mosfet instance Then lt w gt and lt I gt fall inbetween the above minimum and maximum values the corresponding model is picked up from
232. to shrink the integration time step till the minimum value allowed by the finite precision of the computer arithmetic unit Obviously this is time consuming since the length of the integration time step governs the simulation speed To avoid this drawback if the lt trtime gt is specified it defines the minimum integration time step that is lt trtime gt below which the transition can be considered satisfing stvar Integer 0 the number of state variables used in the macromodel MODEL DEFINITION model ModelName vees lt parameter value gt MODEL PARAMETERS macro Boolean false the device is described by an external macromodel module String the name of the shared module implementing the routines of the macromodel evaluate String the name of the routine that evaluates the macromodel args String the name of the routine that handles the arguments passed to the macromodel Linear Voltage Controlled Voltage Source Voltage controlled voltage source is a device with at least 2 terminals the number of terminals must be even and those above the first 2 is any are the controlling terminals There is a lt gain gt parameter for each controlling pair that linearly relates the output voltage to the voltage across the corresponding input port External macromodels are supported The routine describing the macromodel is called with the following ordered arguments lt input gt int an index that uniquely identifies the insta
233. tor The bipolar transistor model is adapted from the integral charge model of Gummel and Poon and includes several high bias level effects The model also includes a collector substrate capacitor that is not installed if the collector and substrate terminals are connected to the same node It defaults to the simpler Ebers Moll model if certain parameters are left unspecified ELEMENT DEFINITION Name collector base emitter substrate thermal ModelName lt parameter value gt ELEMENT PARAMETERS area Junction area in square meters region DC operating region O off 1 on 2 rev 3 sat temp Device junction temperature m Integer 1 number of devices connected in parallel noisetype Integer 7 types of generated noise possible choices are 0x0 noiseless BJT Ox1 thermal noise 0x2 shot noise 0x4 flicker noise The value of this parameter can be the result of the bitwise or of types of noise sources MODEL DEFINITION model ModelName bjt lt parameter value gt MODEL PARAMETERS npn NPN bipolar transistor pnp PNP bipolar transistor type String type of bipolar transistor allowed values are npn or pnp 4 4 semiconductor device 59 is Saturation current js Saturation current bf Forward beta nf Forward emission coefficient vaf Forward early voltage vbf Forward early voltage va Forward early voltage ikf High current corner for forward beta jbf High current corner for forward beta ik Hig
234. tor keyword to access it is lt freqindices gt The DC component is the first element of index 1 in the vector of output spectra Specific beat can is defined by the expression f i j G 1 F1 G 1 F2 where i is the index referring to the first fundamental F1 and j is the index refering to the second fundamental F2 For example by assuming that the name of the spectral analysis is lt Sp gt the index of the second fundamantal can be retrieved inside the body of a control analysis through the statements indices get Sp freqindices i 1 j 2 idx2 indices i j 1 Note that index i and index j have been increased by 1 and the same has been done for idx since alements of the indices matrix are numbered by starting from 1 1 ANALYSIS DEFINITION Name spectral lt parameter value gt ANALYSIS PARAMETERS annotate Integer 5 degree of annotation Values from 1 to 7 turn on displaying of generic informations about the analysis run Values above 7 are bitwise and turn on specific printouts In particular 0x08 print some information about the convergence of the Newton method 0x10 print MNA matrix and RHS vector at each iteration of the Newton 4 9 Advanced frequency domain analyses 141 method 0x20 at each iteration of the Newton method the MNA matrix is writted in a file named as the analysis 0x40 print matrices related to probes This option has meaning only when the probe
235. tor during the analysis It is not ensured that these time instants are effectively met However during the analysis these time instants are considered and the integration time step selection algorithm tries to meet these points savetime Real results are saved in the output file when simulation time is greater than the value of this parameter This parameter can be useful to reduce the amount of results saved in the output file tstart Real time instance at which the transient analysis begins stop Time stop or end of the time domain analysis tstop Time stop or end of the time domain analysis trabstol Real 1 0e 12 absolute convergence tolerance for capacitor voltages and inductor currents trreltol Real 1 0e 3 relative convergence tolerance for capacitor voltages and inductor currents vabstol Absolute voltage tolerance employed by the Newton algorithm at each time point of the time domain analysis default 1 0e 6 vreltol Relative voltage tolerance employed by the Newton algorithm at each time point of the time domain analysis iabstol Absolute current tolerance employed by the Newton algorithm at each time point of 4 6 Basic analyses 113 the time domain analysis default 1 0e 12 ireltol Relative current tolerance employed by the Newton algorithm at each time point of the time domain analysis opiabstol Absolute current tolerance for the evaluation of the initial conditions opireltol Relative current tolerance for the
236. ugh the Newton iterative method that is applied locally to each device partitioner Integer 0 this option turns on the circuit partitioner Allowed values are O the circuit partitioner is kept off 1 the circuit partitioner is turned on The circuit partitioner considers unidirectional elements such as mosfets and partitions the circuit into sub circuits The solution of the sub circuits is organised in a directed graph with no cycles flow graph The flow graph is traversed this means that the sub circuits are solved by starting from the input nodes those that have only outgoing edges to the ending nodes those that have only ingoing edges If circuit partitioning option is not specified it is automatically set equal to 1 when the total number of mosfets of various types plus bjts exceeds a predefined upper limit supernode Boolean true introduce custsets formed by independent voltage sources These cutsets are known as super nodes Super nodes avoid to add the current through the independent voltage source among the unkowns of the problem and thus reduce the size of the problem increasing efficiency subeqns Integer 1000 number of equation to feed each thread circuit partitioning tarjan Boolean false turn on off the EXPERIMENTAL Tarjan DulmageMendelsohn matrix block triangularization algorithm tarjanthresh real integer 1 set Tarjan threshold value if positive real or threshold policy mode if negative integer
237. unction refers to the first and unique driving port There is an other mechanism to access driving currents The circuit element whose current is chosen as driving is identified by the sensedev parameter In this case in the expression the current can be accessed through the function i v1 where vl is the name on the instance used as argument in the sensedev parameter for example in the circuit card Icl 01 02 cccs sensedev v1 func i v1 the current of v1 is accessed through the function i v1 offset Real 0 the offset current generated by the CCCS This current is added to that due to the driving signals Current Controlled Voltage Source Positive current exits the source node and enters the sink node these are the first two nodes of the device There are two ways for sensing current by specifing the sensing nodes or by giving the instance of a sensing device The sensing nodes come always in couple and can be as many as needed In this case a short circuit is placed between the two given nodes Sense current enters sense node and leaves sense node Furthermore a mix of sensing nodes and devices can be specified Each sensing current drives the current controlled voltage source Gains must be specified if the CCVS is not described by means of an external macromodel As many as needed sensing devices can be specified thus the CCVS can be a multi port device External macromodels are supported The rout
238. ures in the same control body Supported data types are scalars vectors matrices and array of matrices Supported types are integer real and complex Available functions abort abort the control analysis abs x arg z computes the argument of z defined as x atan2 imag z real z in radiants 4 7 Simulator control pseudo analyses 119 x conj y computes the conjugate of y If y is real thus x y x cos y computes the cosine of y The y argument can be a scalar a vector or a matrix x cputime get cpu time from the start of the simulator diag eig computes eigenvalues and eigenvectors of a square matrix eye d e derivative func x method delta computes the derivative of the func scalar function with respect to the x variable Different method s can be employed available ones are central which is the default one backward and forward The delta argument defines the initial relative step employed to compute the derivative The estimated value of the derivative is returned as d the absolute error in the estimation is returned in e y db x implemetd the y 20 log10 x function fclose x elaptime get elapsed time from the start of the simulator x fft y computes the fast Fourier transform of y The y argument can be a vector or a matrix If y is a matrix the Fourier transform is performed column by column
239. voffcv noia Flicker noise parameter noib Flicker noise parameter noic Flicker noise parameter em Flicker noise parameter 4 4 8 76 Chapter 4 Reference ef Flicker noise frequency exponent af Flicker noise exponent kf Flicker noise coefficient cluster List list of bsim3v3 models composing the cluster A cluster does not describe a bsim3v3 model itself but groups of already defined bsim3v3 models Each model in the list is characterised by the lt wmin gt lt wmax gt lt Imin gt and lt Imax gt parameters When an instance refers to a cluster of models the list is searched and the current width and length of the instance are compared to those of the models in the list When the width and lenght of the instance fall inbetween the min max values specified in one model of the list it is bound to the instance To summarise clusters are a way to choose models by using width and length of the given instance as a selection criterion bounderror Integer 2 the severity of the error given when lt l gt and or lt w gt are out of bounds lt Imin gt lt Imax gt or lt wmin gt lt wmax gt Possible choises are 0 none error is not given 1 warning simulation continues 2 error simulation is aborted Note that if this parameter is specified it overwrites the value of the corresponding parameter given among the lt options gt of the simulator acm Area calculation method selector hdif ACM Parameter Distance Gate con
240. wise linear source pulse source sinusoidal and exponential ones are meaningfull In a periodic small signal analysis pac analysis only the xacmag and xacphase parameters are used DC current and magnitudes are given in ampere and phase is given in degrees ELEMENT DEFINITION Name sink source isource lt parameter value gt 48 Chapter 4 Reference ELEMENT PARAMETERS dc Real 0 0 DC current idc Real 0 0 DC current stepping Enables or disable the source stepping default enabled type String the current source type freq Frequency of the current source isin Maximum current of the sinusoidal source iphase Real 0 phase degree of the sinusoidal source It is considered only in time domain analyses tdsin Real 0 0 the time interval along which the sinusoidal source is kept off The first period of the source beging exactly after this time interval more rigorously the generated waveform is i t isin sin omega t tdsin vphase exp damping t tdsin damping Real 0 0 the damping factor of the sinusoidal source i t isin sin omega t tdsin vphase exp damping t tdsin magi mag2 Magnitude of desired fundamental and harmonics These is no upper limit to the number of harmonics phi ph2 Phase of desired fundamental and harmonics These is no upper limit to the number of harmonics mag Small signal magnitude phase Small signal phase pacmag Small signal magnitude
241. written only in the output file Direct terminal output can be useful but in general is practical only for small circuits Let us now execute the simulation prompt gt pan Simple_001 pan the output of this command consists in terminal output a log file that is essentially equivalent to the screen output and an output directory named Simple_001 raw that contains results of all simulations in the PAN file plus an index file In general log files are used only for debugging purposes since this is the first example that we are going to see let s look at the output with some detail Logfile 2 2 2 Logfile of Simple netlist Simple_001 log Reading circuit from file lt Simple_001 pan gt Netlist successfully read Building global model declarations Building instance and node declarations Mapping node names Building analysis and option declarations Checking for undefined nets NODE and BRANCH SUMMARY 4 Nodes 5 Total equations DEVICE SUMMARY 14 Chapter 2 A simple PAN walk through capacitor isource resistor vecs vsource eewer Reference node is lt 0 gt KK k kk k k FK k k Fk FK K k 2K K K 2K K K OK K K Dc analysis lt Dc0 gt KK KK KK KK FK K K K K K K K K K K K K K Checking circuit topology for DC analysis Installing lt gground gt resistors Finding the initial guess for the Newton algorithm Start of DC analysis Total power diss

user manual

Contents

Download Pdf Manuals

Related Search

Related Contents