2009 pb dws_multigigabit_models_of_lossy_coupled_lines

MULTIGIGABIT MODELS FOR COUPLED LINES by Piero Belforte

dec. 2009

Method for quick and accurate multigigabit model extraction of

coupled line pairs using TDR measuresand DWS simulations

20/12/2009 1Copyright Piero Belforte

TARGETS:

• Mininum experimental setup effort

• Fast measurements using TDR only

• Minumum number of required measurements

• Maximum efficiency of extracted models

• No numerical stability problems

• Possibly Time-Domain only

20/12/2009 2Copyright Piero Belforte

CHOSEN SOLUTIONS

ONE-PORT only TDR measurements

OTHER PORTS LEFT OPEN

MODEL BEHAVIORAL DESCRIPTION

PWL S-PARAMETERS DESCRIPTION

MAXIMUM 3-POINT PWL

DWS for MODEL SIMULATION

2D SOLVER to get STARTING LOSSLESS MODEL

MODEL PARAMETERS FOUND AS OPTIMUM MATCH WITH MEASURES OF the WHOLE SETUP

20/12/2009 3Copyright Piero Belforte

TEST VEHICULE

• COUPLED MICROSTRIP PAIR

• FR4 SUBSTRATE H=1.4mm

• Length=136mm W=140um S=394um (Center)

• ACCESS THROUGH VIA HOLES and 60um PADS

20/12/2009 4Copyright Piero Belforte

VIA HOLES and PADS

• 60 mils (1520 um)square pad

• 50 mils (1270 um) via diameter

• Dielectric (FR4) height H=1400um

• Very large compared to coupled pair

20/12/2009 5Copyright Piero Belforte

Full Ideal Setup

• No Vias, No TDR parasitics, Ideal lossless lines

• Mode data from 2D Field Solver

• COMMON MODE: TD=891.4 ps

• DIFFERENTIAL MODE: TD=710.9 ps

20/12/2009 6Copyright Piero Belforte

Ideal (Lossless model) from 2D FieldSolver (DWS description)

ModalAdaptors Lossless Modal

TransmissionLines

VoltageEigenVectorsMatrix (2X2)

This model is inserted as first order approximation in the DWS TDR setupdescription .

20/12/2009 7Copyright Piero Belforte

PSEUDO-DIFFERENTIAL SETUP

• TDR connected at port #1

• Remaining ports left open

3

42

1

20/12/2009 8Copyright Piero Belforte

COMMON MODE SETUP

Ports #1 and #2 connected together

Ports #3 and #4 left open

TDR at ports #1 and #2

1

2

3

20/12/2009 9Copyright Piero Belforte

Full Ideal Setup:Simulation Results

Pseudo-differentialsetup

Common mode setup

2xDeltaTD=360ps

20/12/2009 10Copyright Piero Belforte

Ideal Pseudo-diff vs Measure

20/12/2009 11Copyright Piero Belforte

Ideal Common mode vs Measure(Window 6ns)

20/12/2009 12Copyright Piero Belforte

Common mode TDR

Actualmeasure

LosslessModel

The differences between the actual measure and the lossless modelare outlined in yellow

20/12/2009 13Copyright Piero Belforte

Ideal Near-end Xtalk vs Measure(Window 5ns)

20/12/2009 14Copyright Piero Belforte

Adding Vias and parasitics

• Models of Vias and setup parasitics (Launchcable tip and common mode connection) are extracted by means of TDR measures.

• The extracted models are the inserted in the setup simulation model to evaluate theireffect on overall DUT response

20/12/2009 15Copyright Piero Belforte

TDR cable tip model

• The effects of TDR semi-rigid launch cable are not negligible and have to be takeninto account

• The unshielded tip of the cable has a major discontinuity contribution on TDR response

• An accurate model can be derived from an actual TDR measure of a grounded tip

Semi-rigid coax. 1.8mm OuterDiameter.

UnshieldedPTFE dielectric

Center conductor (diameter= .3mm)

Via and pad

Microstrip end

20/12/2009 16Copyright Piero Belforte

Model of TDR Cable tip

• A LCR DWS model of the tip is extracted from the measure of the grounded tip

• The model parameters are optimized to get the best match between measureand simulation

20/12/2009 17Copyright Piero Belforte

Pad and via model

• From the TDR response of an open via a circuital model is extracted by optimization

of the model response

The TDR launch cable tip is also taken intoaccount

20/12/2009 18Copyright Piero Belforte

Optimized model response vs actualmeasure

20/12/2009 19Copyright Piero Belforte

Via and pad DWS optimized modeldescription

• The model is described as DWS subcircuit(stub TL with radiation losses and RC lumped

load to gnd)

20/12/2009 20Copyright Piero Belforte

Adding Vias and parasitics

• The previous models of Vias and setupparasitics (Launch cable tip and common mode connection) are added to the wholesimulation model including the lossless modal

description of the coupled pair.

20/12/2009 21Copyright Piero Belforte

Pseudo Differential TDR: results

Actual TDR Measure

ModelSimulation

Several differences can be still pointed out between this modeland the actual measure

1

2

3

4

20/12/2009 22Copyright Piero Belforte

Adding losses and adjusting DUT model parameters

Losses are added to the DWS simulation model of the DUT. A simplifiedBehavioral model of both even and odd modes is used instead of losslesstrasmission lines.A three point PWL description of S21 is used to model losses. The S11 isassumed ideal. The coordinates of S21 PWL description are optimized to matchthe measurement results . The mode impedances and propagation delay timesare also optimized to match the actual measure. The optimization procedure isapplied to both measurement setups (pseudo- differential and common configurations with all other ports left open).In this way a true de-embedding of coupled line parameters is performedbecause all other setup effects are taken into account in the DWS netlist.To check the results a near-end crosstalk simulation is performed and comparedto the actual measure.

20/12/2009 23Copyright Piero Belforte

Coupled pair lossless DWS model usingBimodal adaptors

20/12/2009 Copyright Piero Belforte 24

Adding setup effects and losses(1)

20/12/2009 Copyright Piero Belforte 25

Adding setup effects and losses (2)

20/12/2009 Copyright Piero Belforte 26

Coupled pair lossy DWS model:optimized results

20/12/2009 Copyright Piero Belforte 27

Checking the near-end crosstalk: setup

20/12/2009 Copyright Piero Belforte 28

Checking the Near-endcrosstalk:model vs measure

20/12/2009 Copyright Piero Belforte 29

TraceAnalyzer: input configuration

20/12/2009 Copyright Piero Belforte 30

TraceAnalyzer: results

20/12/2009 Copyright Piero Belforte 31

Results comparison

Z0e (ohms) Z0o (ohms) TDe (ps) Tdo(ps)

Trace AnalyzerFromcrossectiondata

217 79.87 755 688

DWS meas. basedoptimization

190 64 725 675

20/12/2009 Copyright Piero Belforte 32

De-embedding setup and via effects

21/12/2009 Copyright Piero Belforte 33

De-embedded model: effect of losseson reflected edge rise time

21/12/2009 Copyright Piero Belforte 34

TLineSim results for a single line: effectof losses on the reflected edge

21/12/2009 Copyright Piero Belforte 35