asimple ads schematicfor space mapping · 2agilent ads 2005a, agilent technologies, 1400...

2009 IEEE MTT-S International Microwave Workshop Series onSignal Integrity and High-Speed Interconnects (IMWS2009-R9)

A Simple ADS Schematic for Space MappingQingsha S. Cheng, Member, IEEE, John W. Bandler, Life Fellow, IEEE, and

Slawomir Koziel, Senior Member, IEEE

Simulation Optimization Systems (SOS) Research Laboratory,McMaster University, Hamilton, ON, Canada L8S 4K1, www.sos.mcmaster.ca

School of Science and Engineering, Reykjavik University, Kringlunni 1, IS-103 Reykjavik, Iceland

Abstract-In this paper we present a simplified space mappingimplementation in Agilent ADS. All the space mapping steps areintegrated into one ADS schematic. We also describe a genericsequential optimization implementation method using theApplication Extension Language functions to activate anddeactivate certain design blocks or components to achieve thenecessary automatic iteration and looping. The methodology isdemonstrated through two microwave design examples.Index Terms-space mapping, electromagnetics-based CAD,

circuit design, microwave modeling.

I. INTRODUCTION

Space Mapping (SM) technology [1] addresses the issue ofreducing the time-consuming full-wave electromagnetic (EM)simulations of microwave structures with the help of a fastphysics-based model or surrogate for device modeling andoptimization. Many engineers have adopted SM [ 2 ] [ 3 ].Because SM requires more than one model, implementation isnot straightforward. The first implementation was in OSA901.A comprehensive Matlab-based framework, the SMF system,was introduced in 2005 [4].The purpose of this paper is to introduce a strategy to semi-

or fully-automate the SM process in a simpler way than in [5].We illustrate our strategy solely in Agilent ADS2, a widely-used Electronic Design Automation (EDA) system for high-frequency and microwave modeling and design applications.The strategy allows for easy application expansion andapplication-dependent algorithm adjustments.

II. AUTOMATIC SPACE MAPPING ADS IMPLEMENTATION

In [5] an automated four-schematic implementation of thespace mapping technique is described. The steps for input,

This work was supported in part by the Natural Sciences and EngineeringResearch Council of Canada under Grants RGPIN7239-06, STGP336760-06,and by Bandler Corporation.

Q.S. Cheng and J.W. Bandler are with the Simulation OptimizationSystems Research Laboratory, Department of Electrical and ComputerEngineering, McMaster University, Hamilton, ON, Canada L8S 4K1.

J.W. Bandler is also with Bandler Corporation, Dundas, ON, CanadaL9H 5E7.

S. Koziel is with the School of Science and Engineering, ReykjavikUniversity, Kringlunni 1, IS-103 Reykjavik, Iceland.1 OSA90/hopeTM version 4.0, User's Manual, August 1997, OptimizationSystems Associates Inc. (now Agilent Technologies), P.O. Box 8083, Dundas,Ontario, L9H 5E7, Canada.2Agilent ADS 2005A, Agilent Technologies, 1400 Fountaingrove Parkway,Santa Rosa, CA 95403-1799, USA.

implicit or output SM are implemented as four separateschematics: coarse model optimization, fine model simulation,parameter extraction, and re-optimization of the surrogate.

In our new methodology, we combine the four schematicsinto one schematic. The schematic is partitioned into differentfunctional blocks (Fig. 1). This way, the optimization resultsare saved explicitly in the schematic and no temporary filesare required to transfer data between iterations, except for theexternal fine model simulation. The new schematic design iscapable of calling either ADS Momentum or Sonnet em3 as afine model.The new implementation also facilitates manual operation if

preferred. Simply disabling and enabling appropriate blocksin each step will do the trick.Depending on the application, extra steps to make the

optimization robust can easily be added into the schematics.

Fig 1. An ADS schematic including 4 functional blocks.

III. ROBUST PARAMETER EXTRACTION AND COARSE MODELOPTIMIZATION

We take a few measures to improve the robustness of theparameter extraction and surrogate model optimization. Theparameter extraction (PE) process matches the surrogate to thefine model. In some cases, the traditional gradient-based PEfails. Therefore, depending on the application, we implementvarious two-state PE strategies. Normally, we use the ADSrandom search followed by the Quasi-Newton optimization. Ifthe random search leads to a poor match, we restart our SMalgorithm with a two-stage gradient and/or Quasi-Newtonoptimization.

Surrogate optimizations are frequently performed by theADS Minimax optimization routine. Occasionally the routinemay get trapped in a local minimum, especially after aparameter extraction process that brings the surrogate to a

3emTM Version 11.52, Sonnet Software, Inc., 100 Elwood Davis Road, NorthSyracuse, NY 13212, USA.

978-1-4244-2743-7/09/$25.00 C2009 IEEE

surrogateoptimizabonalgorithm

PE algorithm

coarse model! fine modelsurrogate

35 Guadalajara, Mexico, Feb. 19-20, 2009

Authorized licensed use limited to: McMaster University. Downloaded on April 23, 2009 at 14:46 from IEEE Xplore. Restrictions apply.


frequency-shifted design. We use a specification trackingtechnique to guide the path of surrogate optimization.

Step 1 Connect the specifications to form a piecewise-linear specification function.

Step 2 Match the piecewise-linear specification functionto the initial surrogate response using a discrete search.

Step 3 Shift the specification function back towards theoriginal specification in a few stages. In each stage, aMinimax optimization is applied to the surrogate to satisfy theintermediate specification. In other words, the piecewise-linear specification function guides the surrogate to a designthat satisfies the original specification in steps (stages).The specification tracking technique normally works better

in the first space mapping iteration in which the fine modeland the surrogate are usually misaligned in frequency. Thealgorithm disengages the specification tracking step for thefollow-up iterations.

IV. SPACE MAPPING AND SEQUENTIAL OPTIMIZATION

The space mapping is implemented in 4 major steps as in[5]. Instead of four separate schematics we use only oneschematic that combines all the necessary SM algorithmcomponents. We deactivate appropriate components in theschematic in each step. The fine model is embedded using co-simulation (Momentum) and an external executable (Sonnetem) call from the ADS schematic.A. The Space Mapping Algorithm Steps

Step 1 Multi-stage coarse model optimization. [Deactivatefine model, PE optimization routine, PE goals.] (Usespecification tracking described in Section III, if necessary.)

Step 2 Fine model simulation. [Deactivate coarse model,coarse model optimization routine, coarse model goals, PEoptimization routine, PE goals.]

Step 3 Multi-stage parameter extraction. [Deactivatecoarse model optimization routine, coarse model goals.]

Step 4 Multi-stage re-optimization of the surrogate.[Deactivate fine model, PE optimization routine, PE goal.](Use specification tracking, if necessary.)The steps can be implemented manually. We also designed

an AEL4 script to automate the space mapping algorithm steps.B. General Purpose Sequential Optimization/Simulation ScriptWe have designed a general purpose sequential

optimization/simulation program in AEL to facilitateautomatically connecting several simulations or optimizations.To initialize the script, the user creates the necessarycomponents for all the optimization/simulation steps in anADS schematic. The user creates a list of deactivationcomponent rows in our AEL script. Each iterative stepcorresponds to a component row in which the names of thecomponents to be deactivated are saved. The rows are linkedto become another list called steps. The user specifies theloop starting point and the number of loops. The user startsthe sequential optimization/simulation AEL script below.4Application Extension Language, ADS internal C-like script language.

Step 1 The script reads a row in the steps list anddeactivates corresponding components described in that row.

Step 2 The script runs a simulation (optimization).Step 3 The script updates the optimization variables.Step 4 The script saves the schematic (the resulting

datasets are saved too).Step 5 If the algorithm is not completed, the script points

to the next row in the list and goes to Step 1.The user-defined list of deactivation component rows are

shown as follows://list of cortpo(emts to be deactivated in each stepstep2 = list("component 1", "component 3", ...); st row

step2 = list("component 2", "component 4", ...); d row

step3 = list("component 1", "component 2", ..)O1rd row

steps = list(stepl, step2, step3, ...); //list of deactivation rows

decl loop = 0; Hstarting point of the loop, 0 step 1decl repeat = 0; //number of times to repeat the loopUsers can replace the above list with any other combination

of components to implement a different algorithm. Themethod we show here is much simpler than designing multipleschematics and including their names in the batch list file in[5]. Since all the components in each iteration are placed inthe same schematic, it is easier to make changes to the modeland algorithm. The parameter transfer between steps is easilymaintained through the Update Optimization Valuescommand after each step. Adding new iteration steps is assimple as adding a line of code.We connected Sonnet em and Agilent Momentum as fine

models in our space mapping implementation. To driveSonnet em from within the ADS schematic, we need to savethe predicted design parameters in a file on the fly, which canbe done using the write_str( function [Agilent KnowledgeCenter Example ID: 299744]. The next step (simulate thestructure using the saved design parameters) is easily achievedby calling the Sonnet em command line in a var component.

Agilent Momentum can be called using co-simulation in theschematic view. Design parameters and response data areeasily communicated between layout and schematic design.

V. DESIGN EXAMPLES

We show two examples of how this is actually implemented.Our first example is the open-loop ring resonator bandpassfilter [6] shown in Fig. 2. The design parameters are x =

[L1 L2 L3 L4 S1 S2 g]T mm. Other parameter values are: W=0.6 mm, W1 = 0.4 mm. The fine model is simulated in AgilentMomentum. The design specifications are IS21 . -3 dB for2.8 GHz<. < 3.2 GHz, and S21 < -20 dB for 1.5 GHz<. <2.5 GHz and 3.5 GHz < c < 4.5 GHz. The coarse model isimplemented in Agilent ADS (Fig. 3). Implicit spacemapping is applied. The preassigned parameters (substrateheight and dielectric constant) of the components with thesame hatch pattern change simultaneously in the parameterextraction process.

978-1-4244-2743-7/09/$25.00 ©2009 IEEE 36 Guadalajara, Mexico, Feb. 19-20, 2009



w v

+ - *~~~~~~~~~~~~~~~-

WI~~~~~~~~S1.sL IJ;

7 Output

S2

S2

for input space mapping are the shifts of the designparameters, i.e., 3xi, i= 1, ..., 8. We combine the input andimplicit space mapping in one parameter extraction step.

Open-loop Ring filter S-Parameter SimulationLinear Frequency Sweep

Input

Fig 2. Open-loop ring resonator bandpass filter [6] fine model.MLOCTL1 MLINW=Wm TL2L=L1 -L2-L3/2 m SOBND MACLIN MSO W=W mm

Bend1 CIin3 Bend2 L=L2-L3/2MLIN W-=W1 mm W2-W1 mm W_W MLIN

TL3 ~~~W2=W1 Nm Port2W=W1Lmm _ MCLIN S=S2mm MCLIN TL4WWmm Clinl MGAP L=L3 mm CIin2 W=Wl1mL=L4mm W=W1 mm Gap1 W=W1 mm L=L4 mm

Bend3 L-1-3/2mm S-g mm L=L3/2mm Bend4

W=Wlmm 5 l I Y W=WlmmMSOBNJD l hMSOBNDBend5 MGAPBend6

W=Wl mm MLIN Gap2 W=Wl mmTL5 W=Wlmm _ MLIN

W=Wlmm MACLIN S=gmm TL6L=L4mm CIin4 W=W1 mm

Port MSOBND Wl2=W mm MSOBND L=L4mmMLINBend7

V\/2=Wl mm Bed8 MLOCTL7 WWlm

=2m WWlm LW=Wmm L=L3mm W-WmmL=L1-L2-L3/2 L=L2-L3/2

Fig. 3 Open-loop ring resonator bandpass filter ADS coarse model.

In this example we use one stage surrogate optimization andtwo stage parameter-extraction optimization (Random andQuasi-Newton). The schematic is composed as in Fig. 4. Welist the components to be deactivated.The first step is the surrogate (coarse model) optimization.

The PE, PEJ and PEGoall and PEGoal2 components in theParameter Extraction block are deactivated. The finecomponent is also deactivated.The second step is fine model simulation. S_Opt and PE,

PEJ are deactivated.In the third step (the first stage of parameter extraction),

S Opt, SGoall, SGoal2, SGoal3 and PEJ are deactivated.In the fourth step (the second stage of parameter extraction),

S Opt, SGoall, SGoal2, SGoal3 and PE are deactivated.After two iterations, we obtained the responses shown in

Fig. 5. The final fine model specification error is -0.392 dB.We consider a more complicated example, an interdigital

filter design [ 7]. The fine model, shown in Fig. 6, isimplemented in Sonnet em. The substrate height and dielectricconstant are 15 mils and 9.8 respectively. The shielding cover

height is 75 mils. The cell size is set at 1 mil x 1 mil. Thedesign specifications are IS21 < -30dB for 4.0 GHz<c<4.5GHz and for 5.45 GHz <o<6.0GHz andISiI< -0.1dB for 4.9 GHz< co< 5.3 GHz.The initial Agilent ADS coarse model is shown in Fig. 7.

The space mapping approach we adopt is input and implicitspace mapping. In the interdigital filter case, the lengths ofthe microstrip lines (X1,X2,X3,X4,X7 and x8) and the gaps (X5andx6) are defined as design parameters. The implicit spacemapping explores the preassigned parameters in an attempt tomatch the details of the surrogate to the fine model. For theinterdigital filter, the preassigned parameter set consists ofcr(substrate dielectric constant) and capacitors C1 to C6 addedbetween non-adjacent conductors (cf. Fig. 7). The parameters

Fig. 4. ADS space mapping schematic.

-2(

1'{. -4(

-5(

-6(

0w S-PARAMETERS |

S pa-at

Stat=1 5 GHzStop=4,5 GH2zSte~p=0 05 GHz!

,'Ter ,' Ter..-Num=1 coase mode_-3 .--Nm2.,Z=50 Oh. as3 E1E1 Z=50 Ohmn

-L2-1-2 Er2=Er2= L3=L3 Er-L1=Er LI

1-4=L4 Er L2=Er L2S1=-S1 Er L4=Er L4S2-S2 Er Sl=Er S1g=g Er S2=Er S2Hl=Hl Er g=Er gH2-H2H_L1l=H_L1H_L2=H_L2H_L4=H_L4H_S1=H_SlH_S2= H_S2H_g= H_g

Z=50 Oh. N.m=4-77 Li ~~~~Z=50 Oh.

fineL1=L1L2=L2

ion 1-3=L3

n ~~~~~~~S2-S2g=9W=WW1 =Wl

I(S43)" L5=L5'SP1"~~ ~ ~~~~~6L6=L

ModelType=MW

VAR VR6L1=25 4662 Wo W0.6L2-12 6195 VV=10.4

1-4=3 56046 Lo L=L2--L312SlS10 261749{) L67=L1-_L2-1-312S2=0 0932994 og=0.582939 o

ag(S43)"

HH1=0 635{o Erg=10.2 o

H-1=0 635 MH_L2=0 35 o

H_L2=-0635 o

HH_L4=0 63501oH_S10635H_S2=0 635 oH_g=0,635 1.}Er1=10.2 {o}Er2=10.2 {olErL1=10 2 {olEr L2=10,2 'o}ErL-4=1012 oEr Sl=10 2 oEr S2=I0 2o

1.5 2 2.5 3 3.5 4 4.:frequency [GHfz]

(a)

-0

-20

-3

-40 --

-50

-60

-7

-83 I__ 1___1.5 2 2.5 3 3.5 4 4.:

frequency [GHfz](b)

Fig. 5. Fine (-) and surrogate (---) responses: (a) after initial coarse

model optimization; (b) after 3 iterations.

978-1-4244-2743-7/09/$25.00C2009 IEEE

OT Surrogate Optimization

S OptOptimType=Mimimax OptVarN2-'SZ''Maxlt,ers=25 OptVr3-l"'S'Desire~d"rror0,0 OptVarl41--L4`'StatusL-evel=4 OptVaql5j-'L3''Fin,alAn,alysis="Non,e" Ota,[l'1-12`Nraieol=o OptVr7='L'N-Betalz =ys seIGoals=yes1ta[]-L1SaveSolns=yes SaveCurrentEF=no.Sa-eGoals=yesSaveOptimVars=nUpdateDat.ase=yesSave Nommial=yesSaveAl lteations=noUseAIIOptVars=no.OptVar[l=g

GOL GOAL GOAL

G.al Gc,~a Go.1SGoall SGoal12 SGoal3Expr="db(S21 )" Expr="db(S21 )" Expr="db(S21)Si, .sta,neName="SP1". Sim.nstanceName="SPl" SimlrstanceName-"SP1".Mi=Ms- Min=-3Max-20 Max-20 Max=Weight- Weight- Weight-R:r,geVa[lI=''f,eq' RargeVar[1 ]="fre~q` Ran,geVa[l ]=''freq'RangeMin[1]=15 GI, Ra,geMi,[l]=3.5 GHz RargeMin[1]=2.8 GHz!Ran,geMa[l 1=2.5 GHz! RangeMax[l]=4.5 GHz RargeMax[1]=3.2 GHz:

Parameter ExtracticZ OPTIM | ~~~~~~~GOAL

PE PEGoallOptimType=Ra,d.,, OptVar[2]=`lH2` UseAIIGoals=yes Exp,=r-_m2m -.Maxlt,rs=25 OptVar3j=`'H_S2` Sav,eCurr,ntEF=no. Simlrnsta c,Name=-DesiredError0.0 OptVar[4]=`H_Sl` Mir-l1e-6StatusLevel=4 OptVar[5]=`H_L4` Max=l1e6Fin,alAn,alysis=`None" OptVar[6]=`H_L2` W iht-NormalzeGoals=no OptVar[7]=`H_Ll` Ra,eVa,Se~tBes.tValues=yes OptVar[8]=`H_g` RargeMVat1SaveSolns=yes OptVar[9]=Èr1` RargeMaSa-eGoals=yes. OptVar[l10]=Èr2`S-vOptimVars=n OptVar[1]="Er S2ÙpdateDataset=yes OptVar12]="Er_S1"SaveNmial=yes OptVar1]"E_4 GOALS-vA zertions.=rn OptVa[4]="Er L2ÙseAIIOptVars=no OptVa,r[1l5]="Er Ll" G.aOptVar[l]=`H1"- OptVar[6]="Er_g" PEGoal2

Exp,="imag(S21)-imaSimlr,stan,c,Na-e=".Min=-1e-6

opdm RangeVMin[1]=OptimType=Quasi-Nevton OptVar[2]="H2` UseAIIGoals=yes RanrgeMax[1]Maxite,s 25 OptVar[3]="H_S2" Save~CurrentEF=noDesire)dError=0 0 OptVar[4]=`H_Sl`StatuLevel=4 OptVar[5]=`H_L4"Fin,alAnalysis="None` OptVar[6]="H_L2`N.rnalfizeGoal-.n OptVar[7]="H_Ll`SetBes.tValues)=yes OptVar8=HgSaveSolns=yes OptVar[9]="Erl"Sa-eGoals=yes OptVar[10]=Èr2`SaveOpti,,Va-rs=n OptVar[J1]="Er S2Ùpdate~Dat-st=yes OptVar[12]="Er S1Sa-eNomimal=yes OptVar[l13]="Er L4`S-vAiteations.=rn OptVar[14]="Er L2"Use~AIIOptVars=rno OptVar[15]="Er Ll"OptVar[l1]=`Hl" OptVar[l16]="Er g"

J __000

-4

0n

Is t

.5

-70

.5

37 Guadalajara, Mexico, Feb. 19-20, 2009



Fig. 6. Interdigital filter: fine model structure and dimensions [7]; thedimensions are in mils.

1MLIN

TL17 MLINLW=W il TL21

Fig. 7.xInterdigital filter: coarse modelTinADS.

ADS

C,fF CsfF C,fF LE ADS

T-1-20 /T-V2

2 3W-30L19

L2 ,W

-70,,/ / 0

-60 -7--

4 4.5 5 5.5 6frequency [GHz]

Fig. 8. Interdigital filter initial design using parameters from [7]: theIS1 and IS211 responses of the fine model simulation (-) versus thecoarse model after the first parameter extraction (---).

-0

-20 '- '-

7 -30

-50~~~~~-

-60

-804 4.5 5 5.5 6

frequency [GHz]Fig. 9. The interdigital filter IS111 and IS211 responses of the finemodel final simulation (-) versus the surrogate model (---).

We start our design using the optimal design from [7]. Atwo-stage (Gradient followed by Quasi-Newton) parameterextraction process is carried out. We can see that thespecification is not satisfied after the first iteration, but theparameter extraction using input and implicit SM yields agood match between the coarse and fine models (Fig. 8). Thesurrogate (coarse model) is then optimized to predict a newfine model design. The specification tracking technique isdeployed in the first iteration of surrogate optimization. Injust three SM iterations, a good fine model solution(specification error -0.08dB) is obtained (Fig. 9). Ourtechnique connects the steps or iterations, and repeats the loopautomatically. The step-by-step results are saved after eachstep.

VI. CONCLUSIONS

We present a simple ADS schematic implementation of thespace mapping technique. All the space mapping iterationsteps are implemented in one schematic and can be automatedusing the AEL script. This makes the space mappingtechnology more accessible to microwave engineers. Modelswitching and data exchanges between the steps are facilitatedby the new methodology. A generic sequential ADS programthat facilitates connecting and looping for simulation andoptimization is described. The space mapping algorithm isverified in this sequential ADS optimization schematic. Wedemonstrate our scheme using two examples.

REFERENCES

[1] J.W. Bandler, Q.S. Cheng, S.A. Dakroury, A.S. Mohamed, M.H.Bakr, K. Madsen, and J. S0ndergaard, "Space mapping: the stateof the art," IEEE Trans. Microwave Theory Tech., vol. 52, no. 1,pp. 337-361, Jan. 2004.

[2] J.C. Rautio, "A space mapped model of thick, tightly coupledconductors for planar electromagnetic analysis," IEEEMicrowaveMagazine, vol. 5, no. 3, pp. 62-72, Sept. 2004.

[3] A. Kashi, P. Kumar, M. Caron, and V. K. Devabhaktuni, "Aspace-mapping based CAD methodology for modelingtemperature characteristics of combline resonators," PIERSONLINE, vol. 3, no. 7, pp. 1139-1143, 2007.

[4] S. Koziel and J.W. Bandler, "SMF: a user-friendly softwareengine for space-mapping-based engineering designoptimization,"Int. Symp. Signals, Systems and Electronics, URSIISSSE 2007 (Montreal, Canada, July 2007), pp. 157-160. AlsoBandler Corp., Dundas, ON, Canada, 2006,http://www.bandler.com/SMF/.

[5] Q.S. Cheng and J.W. Bandler, "An automated space mappingframework," Frontiers in Applied ComputationalElectromagnetics FACE 2006 (Victoria, BC, Canada, June 2006).

[6]S. Koziel and J.W. Bandler, "Space mapping with multiplecoarse models for optimization of microwave components," IEEEMicrowave and Wireless Components Letters, vol. 8, no. 1, pp. 1-3, Jan. 2008.

[7] J.W. Bandler, R.M. Biernacki, S.H. Chen, and Y.F. Huang,"Design optimization of interdigital filters using aggressive spacemapping and decomposition," IEEE Trans. Microwave TheoryTech., vol. 45, no. 5, pp. 761-769, May 1997.

978-1-4244-2743-7/09/$25.00C2009 IEEE 38 Guadalajara, Mexico, Feb. 19-20, 2009


asimple ads schematicfor space mapping · 2agilent ads 2005a, agilent technologies, 1400...

Documents