DIGITAL WAVE SIMULATION OF LOSSY LINES FOR MULTI-GIGABIT APPLICATION

1
Abstract Frequency domain Vector Fitting (VF) is a well known
technique to generate circuital models of a spatially discretized
lossy transmission lines from theoretical formulation of losses.
The sub-picosecond time steps required by multi-gigahertz
bandwidths and short transmission lines included in the models,
determine long Spice simulation times. A 100X speedup can be
gained using the Digital Wave Simulator (DWS) instead of
Spice. DWS processes the waves of a Digital Network built up
connecting together scattering blocks (circuit elements, nodes
and S-parameter multi-ports) coming from a Spice-like
description. Being a DSP wave processor instead of a classical
nodal equations solver, DWS is computationally very fast and
numerically stable. Comparisons with commercial simulators like
Microcap11 and CST Cable Studio show a good matching of
results. A further 10-100X simulation speedup is obtained if
Piecewise-Linear Fitting (PWLF) is used to describe the time-
domain behaviors of Scattering Parameters. Single or multiple
cell Behavioral Time Models (BTM) can be extracted by PWLF
from TDR/TDT measurements and processed by DWS fast
convolution algorithms. A setup de-embedding can be
performed by pwl breakpoints optimization to fit actual
measurements. A RG58 coaxial cable is analyzed and its VF-
derived eye-diagrams are compared to PWLF measurement-
derived results. At multi-gigabit rates significant differences,
due to cable physical implementation effects, are observed. The
modeling/simulation alternatives (VF/Spice, VF/DWS and
PWLF/DWS) are compared together and the advantages of
PWLF/DWS in term of simplicity, stability and speed are
highlighted.
Index Terms— Behavioral Time Modeling (BTM), , Digital
Wave Network (DWN), Digital Wave Simulator (DWS), Device
Under Test (DUT), Eye Diagrams (ED), Fast Fourier Transform
(FFT), Inter-Symbol Interference (ISI), Nodal Analysis (NA),
Printed Circuit Board (PCB), Piece-Wise Linear (PWL), Pseudo
Random Bit Sequence (PRBS), Signal Integrity (SI), Time-
Domain Reflectometer (TDR), Time-Domain Transmission
(TDT), Transmission Line (TL), Vector Fitting (VF), Vector
Network Analyzer (VNA), Wave Digital Filter (WDF), Worst
Case Eye Diagram (WCED).
I. DIGITAL WAVE SIMULATION: HISTORICAL
BACKGROUND
The application of digital wave algorithms to high-speed
circuit simulations started at CSELT Labs (Turin, Italy) in the
early '70s [1],[2]. This activity was inspired by prof. Alfred
Fettweis [2] early works on Wave Digital Filters (WDF).
While WDF work was focused on a class of digital filters able
to mimic the behavior of analog filters, the aim of CSELT
researchers was to build up a digital wave model of
interconnects among high-speed integrated circuits. The
digital wave equivalent of a Transmission Line (TL) was the
key element of a modular wave network used in the early
simulation program ETA (Easy Transient Analysis) [1]. The
I/O models of active devices were extracted from TDR
measurements and automatically fitted by simulation to
predefined lumped/distributed (RLC-TL) electrical models.
Measurements, model fitting and simulation programs ran on
the same HP9821 desktop computer as shown in Fig.1 [2]. In
the following decade a new version of the simulator called
APICE (Advanced Program for Interconnect Circuital
Evaluation) was developed with extension to lossy
interconnects [4], crosstalk analysis and Worst Case Eye
Diagram (WCED) calculations [5]. In the following years
started the development of a simulation program for general
topology linear and nonlinear circuits. It was conceived with a
Spice-like netlist syntax and its proprietary algorithms
developed by P. Belforte and G. Guaschino included the
solution of the well known DFL (Delay Free Loop) WDF
topology problem. Multi-port S-parameters blocks, piecewise
-linear description of time-domain behaviors, fast convolution
algorithms and nonlinear/time varying resistors were other
features of SPRINT (Simulation Program of Response of
Integrated Network Transients). The companion application
SIGHTS (Standard Interface for Graphic Handling of
Transient Signals) included simulated waveform plotting and
post-processing. WCED, FFT and graphic PWL extraction.
SPRINT and SIGHTS were produced by HDT (High Design
Technology), the company founded in 1988 by P. Belforte
and G.Guaschino. Signal and Power Integrity analysis
included multi-port nonlinear IBIS-like macromodels of
driver/receivers [6], [7]. A collection of state-of-the art
applications was presented at the HP Digital Design
Symposium in 1993 [7] while an application to Multi-Chip
Modules lossy interconnects is reported in [8]. Starting from
the mid '90s, the HDT tools were included in THRIS
(Telecom Hardware Robustness Inspection System) developed
within a CSELT-HDT cooperation [9]. Simulation of radiated
emissions (EMIR) was added in the mid '90s [10]. After
HDT's end of activity in 2001, the founders developed DWS
(Digital Wave Simulator) [11] and DWV (Digital Wave
Viewer) [12]. In 2012 started a cooperation with Ischematics
[13] for developing the cloud-based version of DWS called
Spicy SWAN. A more detailed story of DWS is available in
reference [14].
Digital Wave Simulation of Lossy Lines for
Multi-Gigabit Applications
Piero Belforte, Member, IEEE

2
Fig.1. CSELT LABS, Turin, year 1975: An early application of first-
generation Digital Wave simulator ETA in the design of a .5Gbps high-speed
Multichip Module. The 500Mhz simulated clock waveforms are plotted as eye
diagrams to evaluate jitter and skews.
II. VECTOR VS PIECEWISE-LINEAR FITTING
TECHNIQUES
The basic concepts of two fitting techniques for system
identification suitable for both signal and power integrity
simulation are discussed. Both methods can deal with both
mono and two-dimensional signal propagation related to
lossy interconnects (cables, traces etc.) and power
distribution planes of printed circuit boards. The traditional
method is based on Vector Fitting (VF) [16], [17], a well
known technique to approximate complex functions of
frequency by a rational polynomial expression in terms of
poles and residues. The second is a full time-domain approach
mainly based on behavioral models (BTM) supported by
DWS [6], [7], [8], [9], [11], [12]. DWS processes the waves
of a Digital Network built up connecting together scattering
blocks (traditional circuit elements, nodes and S-parameter
blocks) coming from a Spice-like description. The two
methods are schematically compared in the flow of Fig. 2.
A. The traditional way: Frequency domain Vector Fitting
and Spice simulation
Vector Fitting modeling flow is shown at the left side of
Fig. 2. In case of mono-dimensional propagation, as applies
to lossy interconnects, the physical line is discretized
according to a chosen spatial pitch  in several equal
segments (unit cells) connected in cascade. The pitch  is
related to the required model bandwidth. Each cell can be
implemented as a ZY-TL circuit where Z is a series
impedance, Y a parallel admittance, both frequency
dependent, and TL is an ideal Transmission Line (TL) as
shown in Fig. 3. Starting from the geometrical and physical
crossection data, TL characteristic impedance Z0 is obtained
from theoretical formulations related to the specific cross-
section. TL delay time is the ratio between the spatial pitch
and the propagation velocity. Z(ω) and Y(ω) are obtained
from the theoretical expressions of skin effect and dielectric
losses versus frequency respectively.
Fig.2 . General modeling/simulation flow: on the left the classical method
based on impedance Vector Fitting , on the right the full time-domain
approach based on S-parameters and DWS processing.
Z() 
Y() 
TL Z0, td 
Z0,td*Delt

Fig.3 . ZY-TL unit cell related to a spatial segmentation pitch 
Using the VECTFIT code [17] applied to theoretical
frequency behaviors, unit-cell poles and zeros can be extracted
and then mapped into RL or RLC circuits leading to a Spice-
like cellular model of the interconnect. This ZY-TL model
can be simulated in time domain using a conventional Nodal
Analysis solver like Spice. A careful choice of Spice version is
needed because the presence of several electrically "short"
Transmission Lines requires both an accurate TL model and
the ability to work at small (picosecond or even
subpicosecond) fixed time step. Only a subset of commercial
Spice versions is able to satisfy the previous conditions, while
other versions could lead to inaccurate results or convergence
problems. MicroCap11 (MC11) by Spectrum Software [19]
is a Spice versions able to deal with this class of circuits.
B. An alternative way: Frequency-domain Vector Fitting
and DWS processing
DWS can be conveniently used to replace MC11 to
simulate the ZY-TL models because it requires typically
1/100 of the simulation time working at the same time step.
This means that sub-picosecond time steps can be used to get
accurate results in short times.

3
C. The unconventional way: Time-domain PWL Fitting and
DWS processing
This method can be applied when the time-domain S-
parameters step responses of the DUT are available. Sources
of such responses are TDR/TDT measurements, analytical
formulas [32], or 2D/3D field solvers. In case of measurement,
a setup de-embedding is needed to extract the required
intrinsic DUT S-parameters. Modern TDR and VNA can
directly perform this task, otherwise a de-embedding loop
procedure can be performed by DWS simulation, as shown in
Fig.2. A Behavioral Time-domain Model (BTM) [7], [8] can
be created directly from the files containing the samples of the
de-embedded S-parameters using the FILE mode description
supported by DWS. A more computationally efficient way is
to use the piecewise linear fitting (PWLF) of S-parameter
behaviors as shown in Fig.1. This task can be done manually
on the plotted waveforms using the MCS (Model Capture
System) facility [7] included in DWV [12], the DWS graphic
postprocessor, online tools, or even automatically on the
imported files coming from measurements or simulations. In
case of spatial segmentation in several BTM cells connected in
cascade (micro-behavioral model) [24], [25], [30] pwl
breakpoint optimization can be performed by comparing the
simulated result of the whole cascade of cells to
measurements or to imported simulations. At the end of this
optimization process a de-embedded model of the DUT is
obtained. A similar procedure can be applied to 2D BTM
models of p.c.b power distribution planes [7]. Thanks to the
fast convolution algorithm included in DWS, the processing
time of PWL BTM models is 10 to 100 times shorter than the
time required using the circuital model.
III. COAXIAL CABLE CASE STUDY
A RG58 coaxial cable was chosen as a case study to
compare together modeling methods and simulation tools.
The results coming this study can be extended to lossy lines
with different geometries like other cables and PCB
interconnects.
Time-domain pulse response of coaxial cables has been
deeply studied and several papers have been written in the past
on this topic. Significant papers on pulse response of coaxial
cable are available in [21],[22]. Several documents regarding
the comparisons between measurements and simulations
related to the RG58 coaxial cable have been written by P.
Belforte and S. Caniggia in recent years [23], [24], [25].
According to these documents, the fully terminated 1.83m
long cable interconnect of Fig.4 has been analyzed. A perfect
cylindrical shape, like that considered by CST Cable Studio,
has been considered for the cable cross section, assuming the
conductors made by solid and homogeneous copper. The
geometrical and electrical parameters used for computing
Z(ω) and Y(ω) taking into account respectively skin effect and
dielectric losses are shown in Fig. 4, where εr is the relative
permittivity and tan δ the loss factor of the dielectric. The
related formulas can be found at chapter 7 of the book [18]. In
order to compare the results, tan is set to the default value
used by CST Cable Studio for the RG58 cable. This value is
8e-4, four times the nominal values for polyethylene (2e-4).
The coaxial cable is segmented adopting a 3cm ZY_TL unit
cell physical length (. The equivalent circuits of Z() and
Y() have been extracted by S. Caniggia by means of the
VECTFIT code [17] using 8 poles each. Two series and
parallel RL networks have been respectively synthesized to
map the 8 poles in frequency domain. TL characteristic
impedance Z0 and delay TD are obtained from the well known
coaxial cable formulas. For a 3cm unit cell length TD resulted
to be 151.7608895659 ps. This delay value will be rounded or
interpolated according to the chosen simulation time step and
DWX option. The DWS netlist of this unit cell is shown on
the right side of Fig.5. A cascade of 61 ZY-TL 3 cm long
cells is used to model the 1.83m long cable.
Fig. 4. RG58 coaxial cable theoretical crossection (CST data) and matched
interconnection schematic.
A. Comparative simulations of the unit cell using Spice and
DWS
Several simulations have been carried out using both
DWS and the evaluation version of MicroCap 11 (MC11)
[19]. These comparative benchmarks have been performed to
compare Digital Wave and Nodal Analysis methods in linear
and nonlinear transmission line situations requiring fixed time
step to get accurate results.
As a significant example of comparison, the 3cm ZY-TL unit
cell related to RG58 coaxial cable has been chosen. To stress
the numerical stability of both simulators a fully mismatched
configuration was selected as benchmark circuit. The
SpicySWAN [13] schematic of this configuration is shown in
Fig.5. A DWS cascadable cell (RG58_3CM) is used to
describe the ZY-TL circuit. Serial adaptors (AS) and
grounded inductors are used to calculate two-port inductors
with the trapezoidal rule [11]. A 2V ideal step generator V0 is
connected to the port P2 of the ZY-TL cell while the port P1 is
open terminated by a 50Gohm resistor R0. The generator V0
stimulates the ZY-TL cell with a 2V step having total rise-
time of 10 femtoseconds. The voltage step propagates along
the circuit and when reaches the port P1 is reflected back by
the open termination. At port P2 the signal is inverted and
reflected back by the near zero impedance of generator V0. In
this way a free oscillation voltage pattern is generated at port
P1. At each oscillation the waveform edges are slowed down
by an increasing dispersion due to cell losses modeled by the
ZY network. Fig. 5 shows the simulated voltage waveform at
port P1 for the first 5ns. To get an accurate result a simulation
time step of 1 femtosecond has been chosen for both
SpicySWAN and MC11. This implies the calculation of 10
Million samples on the 10ns window for both simulators.
Z0=50Ω, tpd=5.058ns/m,
L=1.83m50Ω 50Ω
tr
Vg
Vi Vo
0
2V
Conductor
radius=0.395mm
Shield inner radius
=1.397mm, εr=2.3,
tanδ=0.8x10
-3

4
Fig. 5. ZY-TL unit cell related to a 3cm RG58 coaxial cable segment in a free
oscillation test configuration. DWS cell description on the bottom.
The maximum difference resulted to be about 1 microvolt
peak to peak, confirming a good match between the simulator
responses. The time required for a 10ns simulation on an Intel
Quad-Core i7-2630QM 2.00GHz CPU is about 52 minutes for
MC11 and 35 seconds for DWS.
Fig.6. Simulated voltages at port P1 on a 5ns time window. DWS and
MicroCap11 results at 1Femtosecond time step (solid lines) and differences
(DELTA, dotted line).
B. Effects of simulation time step
The circuit of Fig.4 has been simulated using a cascade of
61 ZY-TL 3cm cells to model the 1.83m long cable. Two
ramp rise-times (tr) of 100ps and 1ps respectively have been
used for the input ramp to evaluate the effect of DWS
simulation time-step choice on the output waveform. The
results are shown in Fig. 7. Time step dependence is more
significant with the faster 1ps rise-time input. A significant
error occurs with a 5ps time step. A slight difference on the
edge foot is still visible between the responses related to 0.5ps
and 1ps time steps. This difference is negligible with the
100ps ramp input. From previous results, a good tradeoff
between simulation speed and accuracy is to choose a 1ps
simulation time step for input rise times in the order of 10ps.
Fig. 7. Cable output voltage related to the schematic of Fig. 4 using the 61
ZY-TL model at 3 different DWS simulation time steps: .5ps (solid line), 1ps
(dash dot line), 5ps (dotted line), 1ps and 100ps rise-time input ramps
C. Effects of spatial segmentation pitch
The effect of line length segmentation on digital bandwidth
has been evaluated in terms of Worst Case Eye Diagram inner
profile. Exploiting the DWS computational speed a
comparison has been carried out between a 61-cell (3 cm unit
length) and a 610-cell ZYTL (3 mm unit length) models in the
configuration of Fig. 4. The chosen input transition time is
17ps (20%-80%) with an erfc shape. The result in terms of
inner WCED contours is shown in Fig.8.
Fig. 8. 20 Gbps WCED inner contours comparison for a 3cm and 3mm ZYTL
cell length. 17ps transition time of erfc input signal. DWS simulation time
step=500 femtosec.
The difference between the two WCEDs is relatively small.
This means that the more computationally efficient 61 ZY-TL
cell model can be used at least up to 20-30Gbps getting a
slightly optimistic result in terms of eye opening.
D. DWS versus CST/Cable Studio comparison
A larger bandwidth eye-diagram comparison between
CST/CS 2014 and DWS was performed using a 40Gbps 255-
bit PRBS (25ps bit-time) input sequence..
In this case the time step was set to 1ps for both simulators
in order to achieve a good result accuracy taking into account
the 10ps (total) linear edges of the input signal. CST CS 2014
has been used with modal model and 40Ghz bandwidth
settings. The eye-diagram comparison is shown in Fig. 9. The
eye diagrams look similar in terms of both opening and ISI
jitter despite the high bit-rate used for this comparison. This
happens because also CST Cable Studio uses a VF-derived
ZY-TL cascade of cells to model the cable. The Y portion that

5
models the dielectric losses is implemented by CST using a
RC instead of the RL network of Fig.5 used with DWS.
Fig. 9. 40 Gbps 255-bit PRBS eye diagrams at cable output: upper CST/CS
2014, lower DWS, 10ps input ramp total transition time, time window=20ns,
Time step=1ps for both simulators.
A remarkable difference between the two simulators is the
CPU time required to get the eye diagrams of Fig.9 starting
from the calculated models. An Intel Quad-Core i7-2630QM
2.00GHz CPU requires 12 min with 4 CPU working in parallel
for CST/CS2014 while about 4 sec with a single CPU
engaged is required by DWS. The observed DWS/CST-
CS2014 speedup factor is 720X.
D. Experimental characterization of an actual cable
A RG58/CU 1.83m long actual cable has been
characterized using the setup shown in Fig. 10. The chosen
sample is manufactured by Tasker company. The internal
conductor is composed by 19x0.18mm stranded tinned copper
wires while the 3mm diameter shield is composed by tinned
copper wires with 98% covering. The dielectric is
polyethylene while the 5mm diameter sheath is made in PVC.
The manufacturer gives also these data: 40milliohm/m DC
resistance, 66% relative propagation velocity, and a 6 values
of insertion loss vs frequency from 50Mhz to 1Ghz. At
50Mhz the attenuation is 9.7 dB/100m while at 1Ghz it is 51.8
dB/100m. The TDR/TDT setup is shown in Fig. 10. In order
to get maximum simplicity and minimum cost, the
launch/termination semi-rigid cables are soldered to the DUT
by means of short (1-2mm) splices. The outer conductors of
the same 60mm long SMA semi-rigid cables and those of the
RG58 cable are all soldered to a common ground plane of a
small metalized board. The 22ps (10%-90%) SD24 head
pulse rise-time is fast enough to get significant measurement
up to about 30Ghz. No particular calibration procedure is
required.
Fig. 10. CSA803C/SD24 TDR/TDT measurement setup for the RG58 CU
TASKER cable (l=1.83m) with details of connecting fixture (lower left) and
RG58 cable tip (lower right).
E. S-parameters de-embedding
An accurate mixed BTM-circuital model of the whole setup
has been developed on the basis of actual TDR response of
the setup alone. This model, shown in Fig.11, includes the
actual SD24 pulse generator waveform , the SMA connectors,
the semi-rigid cables connected to DUT and the soldered
connections. The soldered connections, modeled by 115-
ohm 20ps TLs, acts as markers on the measured waveform
simplifying the measurement of actual cable delay [23].
Fig.11 SpicySWAN schematic of the TDR/TDT setup of Fig.10 used for the
de-embedding of S21 pwl breakpoints
Fig.12 shows the measured S11 waveform on a 50 ns
window. Cable impedance micro-discontinuities due to
crossection variations (Fig. 19) are visible as well the narrow
impedance peaks due to soldered joints. S22 measurement
doesn't coincide with S11 due to the asymmetric effect of these
discontinuities. Fig.12 also shows an example of manually
extracted pwl approximation of S11. High frequency
impedance discontinuities are averaged to limit the number of
breakpoints. The narrow peaks due to reflections at soldered
joints are discarded.

6
Fig. 12. Example of PWL approximation of the measured S11 for the 1.83 m
RG58 CU TASKER coaxial cable (CSA 803C setup of Fig. 10).
The measured S21 rising edge is slowed down by the
setup effects. The S21 behavior de-embedding is performed
starting from the pwl samples taken on the measured
waveform. This first approximated set of breakpoints is used
in the DWS simulative model of the whole setup (Fig. 11).
Breakpoint coordinates are then adjusted by iterations of the
optimization loop shown in Fig. 2. A comparison between the
measured S21 rising edge and the simulated edge including
the setup effects at the end of the fitting process is shown in
Fig. 13. The agreement is very good. The DWS netlist of the
extracted BTM cell and related optimized breakpoints is
shown in Fig.14.
Fig. 13. Measured S21 (edge, solid line) using the setup of Fig.11 vs the
fitted BTM model including the setup effects (dotted line). 50ohm
terminations
Fig.14. DWS netlist of the de-embedded single-cell model of the 1.83m
RG58 coaxial cable
The de-embedded S21 waveform has a faster rising edge
with respect the measured one. A total of 18 breakpoints are
required for S11 and 21 breakpoints for S21. The model of
Fig. 15 is supposed symmetrical for simplicity reasons. The
extracted BTM model has been also simulated configuring the
circuit of Fig.11 with TERM1 node connected to open
termination. The result is compared to the actual one-port
TDR measurement of the open cable. This comparison is
reported in Fig. 15 and shows a good agreement between
measurement and simulation with the exception of a slight
overestimation of reflected edge between 20ns and 23ns.
Fig.15. Measured TDR (solid line) vs simulated waveform (dotted line)
including the setup of the RG58 cable open terminated.
The extracted pwl model of S-parameters is then used in
a circuit simulating an ideal TDR (1ps pulse rise-time) to get
the intrinsic response of the cable. In Fig.16 the 61-cell ZY-
TL theoretical model and the measurement-derived BTM
model of the RG58/CU cable are compared together with and
without setup effects.
Fig.16 Comparison of simulated S11 (upper) and S21 (lower) ZY-TL and
BTM models of the 1.83m RG58/CU cable. Both waveforms with and
without setup effects are shown.
Fig.16 points out significant differences between the two
models. The average magnitude of measurement-derived S11
is about 3 times larger than the value coming up from the
theoretically determined ZY-TL model. The measurement-
derived S21 intrinsic rise-time (20%-80% , without setup) is
55ps, about twice the corresponding 30ps rise-time of the ZY-
TL model. A delay difference of 255ps between the intrinsic
model responses is also shown in Fig.16.
F. Circuital versus Behavioral model comparison
S21 rise-time difference has a substantial impact on high
bit-rate eye-diagrams as shown in Fig.17 for a 255-bit PRBS.
At 20Gbps the eye diagram is about closed using the
measurement-based BTM model and nearly open using the
theory-derived ZY-TL model. The impact of S11 model
differences is not significant at the cable output , but becomes
remarkable in case of not fully terminated or bi-directional
transmission configurations [4].

7
Fig.17. Measurement-based BTM vs ZY-TL simulated eye diagrams for
a 20Gbps PRBS-255 bit sequence of the 1.83m RG58 coaxial cable (DWS).
Input risetime:17ps, erfc waveform.
An Intel Quad-Core i7-2630QM 2.00GHz CPU requires
about 0.4 sec to calculate the 100K samples of the BTM eye
diagram alone while about 30 sec are required for the
simultaneous simulation of both models. A comparative eye
diagram evaluation between RG58 and RG223 coaxial cables
using measurement-derived models is reported in [33].
G. Attenuation versus frequency
Cable attenuation in dB/m has been obtained by
simulating the response of the RG58 models at various
frequencies and compared to manufacturer data in Fig.18.
Fig.18 Comparison of specific cable attenuation in dB/m, 50 MHz to 40 Ghz
(upper) and 50Mhz to 1Ghz (lower).
A fair agreement between BTM model and manufacturer data
(up to 1Ghz) is pointed out, while the attenuation produced by
the ZY-TL model is underestimated. The difference between
BTM and ZY-TL diverges for frequencies higher than 2Ghz
and becomes greater than 10dB beyond 20Ghz. The
differences observed at high frequencies can be explained
considering that the actual cable shown in Fig. 19 is made
using tinned copper stranded conductors. Tin has a resistivity
7 times higher than copper. At 10 Ghz the skin effect depth
for tin is in the order of 2 um and becomes comparable to tin
plating thickness.
Fig.19. a) Actual cross-section of the measured RG58 coaxial cable b)
detailed view of the tinned inner conductors
At higher frequencies (30-40Ghz), the current flows
practically only on the tinned conductor surface. This explain
the attenuation difference considering that tin conductor
resistance is 7 times higher than copper, corresponding to a 17
dB higher attenuation. Other sources of different behaviors
can be conductor stranding and proximity effect not
considered in the ZY-TL model.
IV PERFORMANCE COMPARISON
The matrix depicted in Fig. 20 shows the performance
comparison among the three fitting and simulation
alternatives.
Fig. 20. Relative speedup matrix ( simulator vs fitting method)
DWS simulations run typically 100X faster than MC1, while
the fast convolution applied to process PWLF behavioral time
models achieves a further 10-100X speedup. PWLF
combined with DWS processing offers several advantages
with respect VF combined with NA simulation. It is known
that time-domain S-parameters in matched conditions have
simpler behaviors than the corresponding frequency-domain
impedance behaviors [27]. This means that few pwl (tens)
breakpoints can approximate these behaviors with small
errors. To fit the equivalent impedances in frequency domain
a large number of poles and zeros could be required, leading
to potential numerical stability problems. These stability
problems can be corrected by conditioning in some way the
VF data. On the contrary the well known stability and
passivity of Digital Wave processing [15] doesn't require any
conditioning of PWL description to prevent numerical
instability. Circuital models coming from Vector Fitting,
BTM models and even s-plane/z-plane descriptions can be
mixed to IBIS models of active devices in very complex
networks containing hundreds of thousands scattering
elements [31]. Thanks to a linear growth of simulation time
versus network complexity, these models can be processed by
DWS in times that are more than 4 orders of magnitude
lower than those required by NA simulators. More BTM
modeling applications to p.c.b. interconnects, coupled traces
and p.c.b. power planes are reported in [29] and [30].

8
ACKNOWLEDGEMENT
The author thanks Mr. Spartaco Caniggia for the useful
discussions and for the calculated ZY-TL unit-cell data.
REFERENCES
[1] P. Belforte, U. Colonnelli, G. Guaschino "Use of Equivalent Digital Wave
Networks in the simulation of the interconnects among high-speed logic
devices," Alta Frequenza Vol.11 1976 pp 649-660 (Original version in
Italian ) http://dx.doi.org/10.13140/RG.2.1.4546.2240, DIGITAL
WAVE SIMULATION 1975
[2] P. Belforte, "Early applications of DWS modeling and simulation."
Techn. Rep., Jan 2009, http://dx.doi.org/10.13140/2.1.3845.9684,
EARLY APPLICATIONS OF DW SIMULATION
[3] A. Fettweis, https://www.researchgate.net/profile/Alfred_Fettweis, A.
Fettweis: Research Gate
[4] P. Belforte, B. Bostica, G. Guaschino, "Time-domain simulation of lossy
interconnection using digital wave networks", ISCAS '82 Rome,
http://dx.doi.org/10.13140/RG.2.1.3019.2804, ISCAS 82
[5] P. Belforte, G. Guaschino, "Electrical simulation using digital wave
networks," IASTED '85 Paris,
http://dx.doi.org/10.13140/RG.2.1.4447.9207, IASTED 85
[6] P. Belforte, "SPRINT & SIGHTS Applications," HDT Handbook 1989,
http://dx.doi.org/10.13140/RG.2.1.3792.5605, S&S APPLICATIONS
[7] P.Belforte, HDT (High Design Technology), "A high-performance
environment for modeling and simulation of digital systems," HP High -
Speed Digital Systems Design and Test Symposium 1993,
http://www.hparchive.com/seminar_notes/1993_High-
Performance_Environment_for_Modelling_and_Simulation_of_Digital_
Systems.pdf, HP SEMINAR 1993
[8] P. Belforte, F. Maggioni, J. Torres, "Characterization and modeling of
MCM (Multi-Chip Module) substrates and components" IEEE
Transactions on Computers, June 1993 ,
http://dx.doi.org/10.13140/RG.2.1.3006.1287, IEEE Computers June
1993
[9] P. Belforte, F. Maggioni, G.Guaschino, G.Morelli "A new system for the
evaluation of telecom hardware robustness," EMC'96 Rome, ETW97,
http://dx.doi.org/10.13140/RG.2.1.2678.4488, THRIS ETW97
[10] P. Belforte, F. Maggioni, F.Canavero, alii of CSELT, HP, HDT,
Politecnico di Torino "Predictive vs experimental approach for the
evaluation of global robustness of Digital Apparatus," Hewlett-Packard
Design & Integration Symposium Milan, September 25 1997,
http://dx.doi.org/10.13140/RG.2.1.3464.8804, Predictive vs
Experimental Robustness Evaluation
[11] P. Belforte, G. Guaschino "DWS 8.5: Digital Wave Simulator," User
manual 2015, http://dx.doi.org/10.13140/RG.2.1.1892.0160, , DWS 8.5
[12 ] P. Belforte, G. Guaschino "DWV: Digital Wave Viewer," User
manual 2015 http://dx.doi.org/10.13140/RG.2.1.1300.7529 , DWV
[13] E. Pataky, "Ischematics," https://www.ischematics.com/, Ischematics
SpicySWAN
[14] P. Belforte, "DWS story," http://dx.doi.org/10.13140/2.1.3134.1609,
DWS STORY
[15] A. Fettweis, "Wave Digital Filters: Theory and Practice," Proc. of IEEE,
vol. 74, no. 2, Feb. 1986, pp. 270-327, WDF FETTWEIS
[16] B. Gustavsen, A. Semlyen, “Rational approximation of frequency
domain responses by vector fitting”, IEEE Trans. Power Delivery, vol.
14, July. 1999, pp. 1052-106, VF Gustavsen
[17] Vector Fitting Website, https://www.sintef.no/projectweb/vectfit/ ,
VECTFIT
[18] S. Caniggia, F. Maradei, “Signal Integrity and Radiated Emission,” John
Wiley & Sons Ltd, 2008, CANIGGIA MARADEI
[19] Spectrum Software "Micro-Cap 11.0," http://www.spectrum-
soft.com/index.shtm, MC11
[20] CST , "Cable Studio," https://www.cst.com/Products/CSTCS, Cable
Studio
[21] Lawrence Livermore Radiation Laboratory, UC Berkeley, 1964 "Pulse
response of coaxial cables," http://lss.fnal.gov/archive/other/lbl-cc-2-
1b.pdf, Pulse response coaxial cables
[22] J.H.R. Schrader, "Wireline Equalization using Pulse Width Modulation,"
Twente University Thesis 2007,
http://www.nikhef.nl/pub/services/biblio/theses_pdf/thesis_J_R_Schrade
r.pdf, SCHRADER 2007
[23] P. Belforte, "TDR measurements of a RG58 coaxial cable," Tech. Rep.
2013, http://dx.doi.org/10.13140/2.1.5156.6888, TDR MEAS. OF A
RG58 COAXIAL CABLE
[24] P.Belforte, "RG58 coaxial cable: BTM vs analytical models," Tech. Rep.
2013, http://dx.doi.org/10.13140/2.1.4950.3529 RG58 CABLE: BTM
VS ANALYTICAL MODELS
[25] P.Belforte, S.Caniggia, "Rg58 coaxial cable S-parameters, CST vs DWS
vs Spice," Tech. Rep. 2013, http://dx.doi.org/10.13140/2.1.2197.8406,
RG58 CABLE S-PARAMETERS: CST VS DWS VS SPICE
[26] J. Schutt-Ainé, P. Goh, Y. Mekonnen, J. Tan, F. Al Hawari, P. Liu, W.
Dai "Comparative Study of Convolution and Order Reduction
Techniques for Blackbox Macromodeling Using Scattering Parameters,"
IEEE Trans. Comp Packag Manufact. Techol, vol. 1, no. 10, Oct. 2011,
pp. 1642-1650 CONVOLUTION S-PARAM.
[27] P. Goh, M.F. Ain "Fast S-parameter Convolution for Eye Diagram
Simulations of High-Speed Interconnects," SICASE 190 2013: FAST
CONV. EYE DIAGRAMS
[28] R.B Wu "Fast Eye-Diagram Analysis," National University of Taiwan,
EDAP 2013 , FAST EYE DIAGRAM
[29] P. Belforte, "DWS application videos", Techn. Rep. 2015,
http://dx.doi.org/10.13140/RG.2.1.3558.096, DWS VIDEO PLAYLIST
[30] P. Belforte, "Research Gate Website,"
https://www.researchgate.net/profile/Piero_Belforte, Piero Belforte
Research Gate
[31] P.Belforte, F.Maggioni, "DWS model of a complex multi-layer pcb,"
Tech. Rep 2001, http://dx.doi.org/10.13140/RG.2.1.2467.0242, MULTI-
LAYER PCB 220K
[32] P.Belforte,"Step response of lossy lines," Tech. Rep 2015,
http://dx.doi.org/10.13140/RG.2.1.2287.0882, STEP RESPONSE OF
LOSSY LINES
[33] P.Belforte,"Physical implementation effects on multi-gigabit/sec coaxial
cable performance," Tech. Rep 2015, http://dx.doi.org/
10.13140/RG.2.1.3588.7204 , Coax performance
Piero Belforte Born in Turin in 1947, he
received his Laurea degree in Electronics
Engineering summa cum laude in 1970 from
the Politecnico of Turin. From 1970 to 2000
he worked in CSELT, the Research Center of
Telecom Italia as Head of Switching
Techniques Department and then as Head of
Hardware Qualification Department. His
main activity was focused to state-of-the-art
digital switching systems and qualification
tools for the Telecom network.
In 1975 he started the development of
several generations of high-speed modeling and simulation tools using
innovative DSP algorithms for fast computer simulation of high-speed
electronic systems. These tools have been utilized from the beginning up to
present for the design of both state-of-the-art prototypes and commercial
products. In 1988 he founded and directed the company HDT (High Design
Technology) for the development of state-of.-the art CAE tools for SI/PI/EMC
prediction based on digital wave simulation. In the last years of his activity in
CSELT he also created the hardware quality project THRIS involving several
high tech companies and Universities.. From 2001 to present he continues his
research activity as Independent Researcher. In 2012 he started the
development of Spicy SWAN, cloud-based circuit simulation application in
partnership with Ed Pataky founder of Ischematics. He is author of several
publications and international patents in the field of digital electronics with
reference to digital switching systems and techniques for telecom networks,
high-speed electronics, signal and power integrity, circuital modeling and
simulation, electromagnetic compatibility and test equipment for high
performance digital systems.
Email: piero.belforte@gmail.com
Sites: https://www.linkedin.com/in/pierobelforte
https://www.researchgate.net/profile/Piero_Belforte

DIGITAL WAVE SIMULATION OF LOSSY LINES FOR MULTI-GIGABIT APPLICATION

More Related Content

What's hot (20)

Similar to DIGITAL WAVE SIMULATION OF LOSSY LINES FOR MULTI-GIGABIT APPLICATION (20)

More from Piero Belforte (20)

Recently uploaded (20)

DIGITAL WAVE SIMULATION OF LOSSY LINES FOR MULTI-GIGABIT APPLICATION