S-parameter Simulation and Optimization

ADS 2009 (version 1.0) Copyright Agilent Technologies 2009

Slide 5 - 1

S-parameters are Ratios

Usually given in dB as 20 log of the voltage ratios of the waves at the ports: incident, reflected, or transmitted.

S-parameter (ratios): out / in

S11 - Forward Reflection (input match - impedance) Best viewed on a Smith chart. S22 - Reverse Reflection (output match - impedance) S21 - Forward Transmission (gain or loss) These are easier S12 - Reverse Transmission (leakage or isolation) to understand and
simply plotted.

Results of an S-Parameter Simulation in ADS

S-matrix with all complex values at each frequency point Read the complex reflection coefficient (Gamma) Change the marker readout for Zo Smith chart plots for impedance matching Results are similar to Network Analyzer measurements
Next, ADS data
ADS 2009 (version 1.0) Copyright Agilent Technologies 2009

Slide 5 - 2

Typical S-parameter data in ADS

Transmission: S21 magnitude vs frequency Reflection: S11 Impedance on a Smith Chart

Complete S-matrix with port impedance

Note: Smith marker impedance readout is changed to Zo = 50 ohms. Smith chart basics...
ADS 2009 (version 1.0) Copyright Agilent Technologies 2009

Slide 5 - 3

The Impedance Smith Chart simplified...

This is an impedance chart transformed from rectangular Z. Normalized to 50 ohms, the center = R50+J0 or Zo (perfect match). For S11 or S22 (two-port), you get the complex impedance.
Circles of constant Resistance

SHORT

16.7

50

150

OPEN

25

Lines of constant Reactance (+jx above and -jx below)

50

100

Zo (characteristic impedance) = 50 + j0
More Smith chart...

ADS 2009 (version 1.0) Copyright Agilent Technologies 2009

Slide 5 - 4

The Smith chart in ADS Data Display

Impedance: Z

Z=0+j1

Reflection Coefficient: gamma

Z = real

/ imaginary

0 to +infinity / -infinity to + infinity

ADS marker defaults to: S(1,1) = 0.8/ -65 Z0 * (0.35 - j1.5)

Z=0-j2 Z=0-j1

but can be changed to give Z in ohms.

S-parameter controller...
ADS 2009 (version 1.0) Copyright Agilent Technologies 2009

Slide 5 - 5

S-Parameter Simulation Controller

Default sweep variable = freq
The simulator requires a port termination Num = __

Sweep plan can also be used (see next slide). Either way, simulation data results in an S matrix in the data set for the specified range and points. Next, other tabs
ADS 2009 (version 1.0) Copyright Agilent Technologies 2009

Slide 5 - 6

Parameters and Noise tabs

Parameters
Calculate other parameters. Enable Frequency Conversion for ADS system mixer only.

Noise

Turn on for SS noise. If more than 2 ports, specify port numbers for 2 port NF. If not, leave blank.

Dynamic range: Leave blank and get all values.

Next, Sweep Plan...

ADS 2009 (version 1.0) Copyright Agilent Technologies 2009

Slide 5 - 7

Sweep Plan with S-parameter simulations

Sweep Plan is for sweeping FREQ. Otherwise, use a Parameter Sweep for variables (Vcc, pwr, etc.)
Mixer designers: Here is a plan for an RF, LO, and IF.

These are ignored if Sweep plan is selected!

You can also have Sweeps within Sweeps.

Next, Measurement Equations ...

ADS 2009 (version 1.0) Copyright Agilent Technologies 2009

Slide 5 - 8

S-Parameter measurement equations

Simulation palettes have specific measurement equations - you set the arguments. Here, S is the matrix, 30 is the value in dB, and 51 points used to draw the circle.
Example: 3 circles for 3 different values of gain.

Arguments explained briefly here.

You will use some of these in the labs...

ADS 2009 (version 1.0) Copyright Agilent Technologies 2009

Next, matching...
Slide 5 - 9

S-Parameter measurement equations

Simulation palettes have specific measurement equations - you set the arguments. Here, S is the matrix, 30 is the value in dB, and 51 points used to draw the circle.
Example: 3 circles for 3 different values of gain.

Arguments explained briefly here.

You will use some of these in the labs...

ADS 2009 (version 1.0) Copyright Agilent Technologies 2009

Next, matching...
Slide 5 - 10

Creating Matching Networks

Various topologies can be used: L, C, R Avoid unwanted oscillations (L-C series/parallel) Yield can be a factor in topology (sensitivity) Use the fewest components (cost + efficient) Sweep or tune component values to see S-parameters Optimization: use to meet S-parameter specs (goals)

In the lab, you will optimize the match for the amplifier.
ADS 2009 (version 1.0) Copyright Agilent Technologies 2009

NOTE: For a mixer, match S11 @ RF and S22 @ IF.

Use the Smith chart for matching

Slide 5 - 11

Matching means: Moving toward the center of the Smith Chart!

Next, Smith Chart Utility for matching networks...

ADS 2009 (version 1.0) Copyright Agilent Technologies 2009

Paral

lel R

Slide 5 - 12

Smith Chart Utility for matching...

Insert the component in schematic. Select: Tools > Smith Chart Set freq, source Z, load Z. Select components: L, C, R, etc. View responsebuild the circuit.

Next, optimization...
ADS 2009 (version 1.0) Copyright Agilent Technologies 2009

Slide 5 - 13

ADS Optimization Basics

Start with a simulation that gives you results. Set up the optimization which includes:
An optimizer type and search method. A specific goal or specification to be met. Enabled components or parameters to be adjusted.

NOTE: ADS has both continuous and discrete optimization. Yield analysis or a yield optimization is also available.
ADS Optimization in schematic

ADS 2009 (version 1.0) Copyright Agilent Technologies 2009

Slide 5 - 14

Four elements for Optimization setup

1 - Optim controller: set the type, etc. 2 - Goal statement: use valid measurement equation or dataset expression. 3 - Enable component {o} for optimization. 4 - Simulation Controller
Types of ADS optimization...
ADS 2009 (version 1.0) Copyright Agilent Technologies 2009

Slide 5 - 15

ADS Optimization Types Available

= Most commonly used types.

Optimizer Search Method Random random Gradient gradient Random Minimax random Gradient Minimax gradient Quasi-Newton quasi-Newton Least Pth quasi-Newton Minimax mini-max Random Max random Hybrid random/qNetwon Discrete discrete Genetic genetic Simulated Annealing SA
Hybrid is a combination of Random and Gradient.
ADS 2009 (version 1.0) Copyright Agilent Technologies 2009

Error Function Formulation least-squares L2 least-squares L2 minimaxL1 MML1 minimaxL1 MML1 least-squares L2 least P-th seqLP mini-max MM worst case negL2 least_squares L2 least-squares L2 least-squares L2 least-squares L2
NOTE: Sensitivity analysis is available in the optimization controller.
Error Function...

Slide 5 - 16

Error Function Formulation ...

Optimizer Type determines the Error Form:

Least Squares: Each residual is squared and all terms are then summed. The sum of the squares is averaged over frequency. Negated Least-Squares: drives values to their extreme effectively maximizes the error function. The goal is to find a worst typical response for a given set of parameters.

Minimax: attempts to minimize the largest of the residuals. This tends to result in equal ripple responses . Minimax L1: is similar but cannot be less than zero, so it accounts for the most severely violated cases.
Least Pth: The Least Pth EF formulation is similar to L2, except that instead of squaring the residuals, it raises them to the Pth power with P=2 , 4 ,6 etc.

Next, how the EF works with your goals...

ADS 2009 (version 1.0) Copyright Agilent Technologies 2009

Slide 5 - 17

Goals and Error Function

The goals are minimum or maximum target values. The error function is based on the goal(s). The weighting factor prioritizes multiple goals.
Error function is defined as a summation of residuals. A residual ri may be defined as:

W | m - si |
i i

Next, search method examples...

si is the simulated ith response (example: S21= 9.5dB) mi is the desired response for the ith measurement (example: S21=10dB) Wi is the weighting factor for multiple goals: higher number is greater.
NOTE: You can set all goals to be equally weighted.

Simulations continue until the maximum iterations is reached or the error function (summation of the residuals) reaches zero (same as 10 dB).
ADS 2009 (version 1.0) Copyright Agilent Technologies 2009

Slide 5 - 18

Search Method examples

Random analysis often gets you close to the goal (minimum error function).
Error function

Gradient analysis may get stuck in a local minimum (not optimal error function).

Parameter value

Using both RANDOM and GRADIENT can reach the desired goal or, in some cases, a hybrid type such as Genetic. Next, the setup...

ADS 2009 (version 1.0) Copyright Agilent Technologies 2009

Slide 5 - 19

Optimization Controller setup

Setup tab: Select type and set iterations. Default setting use all Goals and VARs or select specific ones in OptVar tab. Parameters tab: type, iterations, etc. All are displayed by default.

Final Analysis: SP1SimInstanceName

Avoid saving unwanted data.

Normalizes multiple goals!

ADS 2009 (version 1.0) Copyright Agilent Technologies 2009

Slide 5 - 20

Enabling components for Opt or Stats (yield)

OPT: Enable and specify continuous or discrete (stepped) variation.

Discrete or Continuous

PPT is an optimization within a Yield Analysis only. Allows value to be shifted to achieve goal.

Use: Options > Preferences > Component Text tab to set the display to show full opt range instead of {o}. Also, noopt or {-o} = disabled after the optimization.
NOTE: DOE is an advanced topic (statistical yield analysis).
ADS 2009 (version 1.0) Copyright Agilent Technologies 2009

Discrete lib parts...

Slide 5 - 21

ADS Discrete Optimization for Library Parts

Inserted library part with listed range of values (like a DAC)

NOTE: For al list of all optimization variables:

Next, Yield
ADS 2009 (version 1.0) Copyright Agilent Technologies 2009

Slide 5 - 22

Yield Analysis: % meeting specs!

Example: 200-400 MHz (50-to-100 ohm) Impedance Transformer. Variables have been optimized {o} and statistical Yield will now test the % of circuits meeting spec. Note that component values (VAR) have defined statistical distributions {s}.

Random seed

Optional Exercise:
ADS 2009 (version 1.0) Copyright Agilent Technologies 2009

Slide 5 - 23

...Yield Analysis Results (data)

NOTE on Trace Options: Uncheck Line Color to get one color.

80% will meet spec Spec is -18 dB S11 from 200 to 400 MHz.
Zoomed in view: results outside of spec.

Yield Dataset results are stochastic.

ADS 2009 (version 1.0) Copyright Agilent Technologies 2009

More results...

Slide 5 - 24

more Yield Analysis results

The example has Several Data Display windows: Histograms, Sensitivity and Sigma plots. Various methods are used and equations/functions are described for both histograms and sigma plots.

MC...

ADS 2009 (version 1.0) Copyright Agilent Technologies 2009

Slide 5 - 25

Monte Carlo Analysis

Similar to Yield but no spec is required. Enabled parameters are varied and you see the response. Also, Process or Mismatch can be selected to see the effects.

Process and Mismatch used for Dynamic Link to Cadence.

NOTE: Click on the parameter to see highlighted range or set to Full in Options > Preferences.
ADS 2009 (version 1.0) Copyright Agilent Technologies 2009

S-probe...
Slide 5 - 26

S- Probes for Impedance

Get impedance (reflection coefficient) for biased circuits by simply connecting the S probe in series.
Probe components palette:

Matching networks

Insert the SP_Probe between components or stages. Set up the simulation at the higher level.

S-probe pair

Data Display template gives Impedance and stability.

ADS 2009 (version 1.0) Copyright Agilent Technologies 2009

Slide 5 - 27

S_Param Test Lab with Test Benches

Examples/Tutorial: TestLab_HOWTO_prj. Shows how to Optimize a two-port model using measured two-port data. Goals are MeasEqns = Difference between modeled and measured Test Bench values goes to zero when optimized!
NOTE: The sequencer can also be used for an upper level design to run simulations in order.

NOTE: Test benches are created using the test bench symbol (Design Parameters).
ADS 2009 (version 1.0) Copyright Agilent Technologies 2009

Slide 5 - 28

Test Lab example (continued)

Hierarchy of model test bench: TB2

Enabled (opt) variables are used for all values in the lower level design. When the optimization completes successfully, TB2 will be a model of the measured device. TB1 is file based and reads-in the S-param values during the optimization process.
Measured test bench: TB1

Next, Ref Nets!

ADS 2009 (version 1.0) Copyright Agilent Technologies 2009

Slide 5 - 29

Ref Nets are Terminations

RefNets (2 types) are terminations that reference the impedance of other schematics in your project: test benches or designs.

Next, frequency sensitive components Term1 has the impedance of the test bench ckt_1 looking into its port 1 and Term 2 has the impedance the design ckt_2 looking into its port number 1.

ADS 2009 (version 1.0) Copyright Agilent Technologies 2009

Slide 5 - 30

ADS allows frequency sensitive components!

Set a Term Z to change with frequency using IF THEN ELSE.

Use an SNP Eqn: Z changes with changing frequency. if then elseif then elseif then else endif

CAPQ and INDQ: Equation describes changing L or C with frequency.

Try these as an extra exercise after the lab!

ADS 2009 (version 1.0) Copyright Agilent Technologies 2009

Slide 5 - 31

Lab 5:

S-parameter Simulation and Optimization

ADS 2009 (version 1.0) Copyright Agilent Technologies 2009

Slide 5 - 32

Steps in the Design Process

Design the RF sys behavioral model receiver Test conversion gain, spectrum, etc. Start amp_1900 design subckt parasitics Simulate amp DC conditions & bias network Simulate amp AC response - verify gain Test amp noise contributions tune parameters Simulate amp S-parameter response Create a matching topology Optimize the amp in & out matching networks Filter design lumped 200MHz LPF Filter design microstrip 1900 MHz BPF Transient and Momentum filter analysis Amp spectrum, delivered power, Zin - HB Test amp comp, distortion, two-tone, TOI CE basics for spectrum and baseband CE for amp_1900 with GSM source Replace amp and filters in rf_sys receiver Test conversion gain, NF, swept LO power Final CDMA system test CE with fancy DDS Co-simulation of behavioral system

You are here:

ADS 2009 (version 1.0) Copyright Agilent Technologies 2009

Slide 5 - 33

First, simulate with ideal components

Plot the S parameter data Write eqn IF-THEN-ELSE for a Term Z. Simulate and list the data

ADS 2009 (version 1.0) Copyright Agilent Technologies 2009

Slide 5 - 34

Calculate C and L values and re-simulate

Reactance of 10 pF at 1.9 GHz and a list of L values:

ADS 2009 (version 1.0) Copyright Agilent Technologies 2009

Slide 5 - 35

Smith Chart Utility for a matching network...

Smart component in schematic becomes a sub-circuit when built:

Start the tool, define value of Z Load at 1900 MHz, insert the component Select the desired palette components (C and L) to move to Smith Chart center Build the matching circuit with one click Next, Optimize!
ADS 2009 (version 1.0) Copyright Agilent Technologies 2009

Slide 5 - 36

Add output matching components and Optimize !

Use Random with 2 Goals: Max for S11 and S22 = -10 dB

NOTE: opt function or {o} indicates components are enabled.

Note: If GOAL RangeVar is not set, the optimizer uses all the simulation frequency points.
ADS 2009 (version 1.0) Copyright Agilent Technologies 2009

Slide 5 - 37

After a successful optimization...

EF = O = goals met! Update the values and disable the Opt components: {-o} = noopt function.

Nominal Optimization found initial Error Function (EF) to be zero or less. Optimization specification is currently satisfied Try to tighten specification

Simulator will tell you if it can improve results!

Lab exercise has final component values for these S-parameters which are good!

ADS 2009 (version 1.0) Copyright Agilent Technologies 2009

Slide 5 - 38

Also, gain & noise circles + stability

S-parameter simulation with gain and noise circles, and stability. These pre-defined measurement equations use ADS functions.
You set the arguments if necessary!

NOTE:You must turn on Noise in the simulation controller for NsCircle and list values.

Circle center is source match for optimum NF.

ADS 2009 (version 1.0) Copyright Agilent Technologies 2009

Slide 5 - 39

Optional: Read / Write data files

Or, try the yield analysis example!

Data File Tool: write an ADS S-parameter dataset as a Touchstone file, then Read it back in... as if it came from a Network Analyzer!

Start the lab now!

Specify the dataset you want to write out. Simulate with a Sweep Plan and compare the results.

ADS 2009 (version 1.0) Copyright Agilent Technologies 2009

Slide 5 - 40