Microsoft PowerPoint - DESIGNING HS CABLE ASSY REVISED
0606_SM.pptDesigning with Samtec High Speed Cable Assemblies
Designing with Samtec High Speed Cable Assemblies
Samtec High Speed Cable Assemblies Samtec High Speed Cable
Assemblies • Agenda
– Background – Why High Speed Cable is Needed – How to Select High
Speed Cable – Modeling/Simulation Data – Electrical Test Data –
Performance Demonstrations – Correlation between Simulation and
Electrical Test Data – Samtec’s High Speed Cable Assembly product
offering – Using Samtec’s Cable Builder and Cable Calculator –
Samtec’s Signal Integrity Group
Why High Speed Cable is NeededWhy High Speed Cable is Needed •
Signal
– Lower cross talk – Lower EMI emission and susceptibility – Higher
frequency and less loss – Lower voltage drop
• Mechanical – Challenging mechanical routing – If design is less
than 10”, consider flex circuits
Length vs. FrequencyLength vs. Frequency • Single-ended • Flex
Assembly, Coaxial Cable Assembly
Fr eq
ue nc
y in
G H
• Things to consider: – Attenuation – DK – Insertion Loss – DCR –
Voltage Drop
Cable Selection (Attenuation)Cable Selection (Attenuation)
0 2 4 6 8
10 12 14 16 18 20 22 24 26 28
0.25 0.5 1 1.5
10 12 14 16 18 20 22 24 26 28
0.25 0.5 1 1.5
Fr eq
ue nc
GH z
38 AWG OD = 0.025” 34 AWG OD = 0.04” 30 AWG OD = 0.05” 28 AWG OD =
0.065” 26 AWG OD = 0.075”
Attenuation @ -3dB, Z = 50 or 100 DK = 2.1Attenuation @ -3dB, Z =
50 or 100 DK = 2.1
• Lower DK allows less insulation – Lower tanΔ due to high % of air
in dielectric
• Less insulation allows smaller OD: – Can use a larger center
conductor – Less insertion loss (due to lower tanΔ)
• Lower DK = more flexibility in length and bandwidth
0.0750 (1.90)0.0750 (1.90)
0.0650 (1.65)0.0650 (1.65)
Cable Selection (Insertion Loss Attenuation)Cable Selection
(Insertion Loss Attenuation) • Based on Insertion Loss (dB at
frequency) at a specific length
– At a fixed conductor size – If coax / twinax, size is a BIG
concern – By decreasing DK, then frequency would increase at a
given loss
DK of 2.1, -5dB @ 1.2GHz, C= 29 pF/FT
DK of 1.4, -2dB @ 1.2GHz, C=23 pF/FT
Cable Selection (DC Resistance-Voltage Drop)Cable Selection (DC
Resistance-Voltage Drop)
0.00000
0.10000
0.20000
0.30000
0.40000
0.50000
0.60000
0.70000
DCR of Conductors (AWG) Voltage Drop
0.000 0.010 0.020 0.030 0.040 0.050 0.060 0.070
22 24 26 28 30 32 34 36 38
Calculated DCR 12”
Voltage Drop at 12” for various AWG at specific current
.05 Amp
.1 Amp
• Based on Voltage Drop or Ampacity – Increasing Conductor Size
AWG
Inside and Outside the BoxInside and Outside the Box
RIBBON TWINAXRIBBON TWINAX
RIBBON COAXRIBBON COAX
ModelingModeling • Complex electromagnetic modeling
is performed on Samtec cables – PCB transitions modeled with
CST
Microwave Studio • Accounts for structural discontinuities in
the transition assembly – 2D frequency dependent modeling of
cable performed with Ansoft 2D • Accounts for frequency
dependent
characteristics of the cable
interconnect elements is necessary for accurate simulation of
systems
– Size matters • The longer an element is, the more important that
accurate
frequency dependent modeling is performed – Traces, long
connectors, flex, cables ….
» For short, well controlled elements, such as short board-
to-board connectors, losses may be ignored with low error
– Irregularity matters • Irregular and 3-dimensional objects
generally have non-TEM
propagation modes and require modeling in the frequency
domain
– Non-uniform traces, vias, SMA launches, connector transitions,
cable transitions, connector breakout regions, antipads ….
TEM Modeling of Uniform StructuresTEM Modeling of Uniform
Structures • Uniform long structures may generally be modeled
using
TEM or Quasi-TEM assumptions with 2-D field solvers – Traces, coax,
some connector cross sections ….
• But error increases if the field solver does not model frequency
dependent conductor and dielectric losses correctly – Most do not!
– Finite field penetration into conductors (skin effect) is
often
only partially modeled. Usually the resistive portion of skin
effect is calculated, while the inductive portion is ignored
» Most solvers provide one value for inductance, which is
incorrect!
Skin DepthSkin Depth • At high frequencies, most of the current
flows
on the surface of conductors • For copper, the skin depth defines
the point
at which the magnitude of the field has been reduced by 1/e or 37
percent
MHzin fmicrons,in δ
copperfor , f 1x microns66δ =
Coax Center Conductor Current Distribution - 38 Gauge Coax (4 mil
diameter) Coax Center Conductor Current Distribution - 38 Gauge
Coax (4 mil diameter)
1 MHz (66 micron) 10 MHz (21 micron) 100 MHz (6.6 micron)
1 GHz (2 micron) 3.5 GHz (1.1 micron)
Frequency Dependence of ResistanceFrequency Dependence of
Resistance
DC resistance
AC resistance
Actual resistance values will be strongly dependent upon the
conductor cross-section.
Modeled with Ansoft Maxwell 2D
Frequency Dependent Inductance Frequency Dependent Inductance
Modeled with Ansoft Maxwell 2D
External Inductance Asymptote
High frequency surface inductance limit.
Inductance at 350 ps risetime
Inductance at 35 ps risetime
Inductance at 150 ps risetime
Variation of inductance in normal operating region is 2% to 3% of
extracted value at infinity.
Final Inch® Trace ModelingFinal Inch® Trace Modeling • Our approach
to trace modeling
– Utilize Ansoft Maxwell 2D • Finite element quasi-static field
solver • Capable of extracting frequency dependent R and L
– Measure (when possible) substrate material properties across
frequency (Er and Loss tangent) and use during
parameterization
– Extract trace parameters using a parametric sweep • Sweep from 10
Hz to 50 GHz for accuracy across all
frequency bands • Utilize Z and Y matrices
– RLCG matrices do not include losses in Ansoft 2D
– Create HSPICE W-element table model • Automated process to
extract Z and Y matrices to create
compatible table model
Snippet of Final W-element Table ModelSnippet of Final W-element
Table Model .MODEL final_inch_se W MODELTYPE=table N=1 + RMODEL =
final_inch_se_R LMODEL = final_inch_se_L + GMODEL = final_inch_se_G
CMODEL = final_inch_se_C
* ###R-model### * data type = * R-model .MODEL final_inch_se_R SP
N=1 SPACING=nonuniform VALTYPE=real + INTERPOLATION=spline +
DATA=32 * ============= ============= ============= * FREQUENCY: +
0.0000000000000000e+000 * TABLE ELEMENTS: * === row 1 === +
5.1907890527286469e+000 * ============= ============= =============
* FREQUENCY: + 1.0000000000000000e+002 * TABLE ELEMENTS: * === row
1 === + 5.1907890900627756e+000
TEM vs. Non-TEM Modeling Of Non-Uniform Structures TEM vs. Non-TEM
Modeling Of Non-Uniform Structures • Non-uniform structures require
modeling with a 2.5-D
or 3-D full wave approach – Fields generally do not meet TEM or
Quasi-TEM
assumptions • Electric and Magnetic fields are not reasonably
orthogonal • Lumped and/or distributed model approximations are
no
longer accurate – Network parameters (S-parameters) are generally
the
best way to model the broadband performance of these
structures
• Full wave field solvers and simulators like CST Microwave Studio
can be used for the extraction of these structures
Simulation ModelSimulation Model • EQCD circuit Simulation,
1.0Gbps, 250ps, 6in coax length •
*******************************************************************************
• * Copyright 2003, Samtec, Inc. • * •
*******************************************************************************
• *!!!!!!!!!! USER INPUT REQUIRED HERE !!!!!!!!!! • * Specify
length of coax in inches • .param coax_length_inch = 6 Cable Length
• ************************************************ • * convert
length to meters • .param coax_length = '(coax_length_inch) *
0.0254'
• .inc './mod/Mqte1qse1.sp‘ Samtec Connector Model
• .inc './inc/Stimulus_se_1bit_prbs3_56rec_1.0Gbps_250ps_50ohm.inc'
• .inc 'EQCD-DV_20line_circuit.inc‘ Samtec Cable Model
Correlation to MeasurementCorrelation to Measurement
Measured vs. Modeled IL of 12" Hitachi Coaxial Cable (Vert Scale
Changed)
-2
-1.8
-1.6
-1.4
-1.2
-1
-0.8
-0.6
-0.4
-0.2
0
0.00E+00 2.00E+08 4.00E+08 6.00E+08 8.00E+08 1.00E+09
Frequency (Hz)
IL (d
Measured
Modeled
Measurement BoardsMeasurement Boards
Cable AssemblyCable Assembly
Transition Close-upTransition Close-up
• DC Drop
Family of Insertion LossFamily of Insertion Loss
Return LossReturn Loss • Return loss is energy lost due to
reflections in the cable
or assembly • Caused by mismatch in impedance
– In the cable • Design • Manufacturing (variance from design,
repetitive features) • Handling and installation (kinks)
• Variation due to structural discontinuities is called structural
return loss
– Structural return loss happens in: • Transition • Connector •
Cable
Return LossReturn Loss
SkewSkew
Skew of cable assemblies should be carefully considered for each
application.
Twinax should be used when intra-pair skew of differential pairs is
critical.
CrosstalkCrosstalk
Crosstalk of cable assemblies is primarily created within the
transition structures.
QTE/QSE 10 Gbps Serdes Demonstration BoardQTE/QSE 10 Gbps Serdes
Demonstration Board
Bi-directional transmit and receive boards shown with twinax cable
attached
Final Inch® and 1-Meter EQCD Coax, with Accelerant Networks AN6425
PAM-4 Serdes Final Inch® and 1-Meter EQCD Coax, with Accelerant
Networks AN6425 PAM-4 Serdes
PAM-4 eye pattern for Accelerant Networks AN6425 at 6.22 Gbps with
Samtec QSE/QTE Final Inch® test board and a 1-meter long 38 AWG
EQCD micro coax assembly showing excellent eye opening
Boards With 6” Coax With and Without Equalization Boards With 6”
Coax With and Without Equalization
PRBS pattern as transmitted through connectors, 6” cable
and PCB only
Equalized
Boards With 0.5 Meter Coax With and Without Equalization Boards
With 0.5 Meter Coax With and Without Equalization
Binary eye pattern for MAX3952 PRBS, PCB trace, QTE/QSE connectors,
0.5- meter long 38 AWG EQCD micro coax and MAX3805 adaptive
equalizer shows excellent eye opening
1 Meter Coax With and Without Equalization1 Meter Coax With and
Without Equalization
Binary eye pattern for MAX3952 PRBS, PCB trace, QTE/QSE connectors,
1-meter long 38 AWG EQCD micro coax and MAX3805 adaptive
equalizer
2 Meter Twinax With and Without Equalization2 Meter Twinax With and
Without Equalization
PRBS pattern as transmitted through connectors, 2 m twinax and PCB
with equalization
Samtec’s Transmission Line OfferingSamtec’s Transmission Line
Offering Gauges In Use • 50
– 38 AWG coax – 34 AWG coax – 30 AWG coax – 26 AWG coax
• 100 – 30 AWG twinax – 26 AWG twinax
• Cable produced by Samtec-owned cable manufacturing facility
Terabit will be integrated into this available cable offering next
quarter
Samtec Data Rate CablesSamtec Data Rate Cables • End-to-End
solutions with in-house design and
manufacture of: – High Speed Connectors – High Speed Cable – SI
Modeling and Support – Qualification Testing
• “Mix-and-Match” capabilities that include – An extensive line of
connectors
• Standard and custom
– A wide variety of cables • Coax and twinax
Signal Integrity SupportSignal Integrity Support Signal Integrity
Division
• SI Center: Comprehensive, user-friendly web site for High Speed
connector SPICE models, reports, performance data, drawings,
searches
• SI Group: Personal, “live person” EE customer interface for
application issues
• SI Services: Custom design, modeling, testing of circuits,
subsystems, or complete systems
SummarySummary High Speed cable assemblies are a strategic product
for Samtec
– Complements High Speed connector offering
– Leverages Samtec’s in-house testing and modeling
capabilities
– Samtec is investing significant financial and personnel
resources
– High Speed cable solutions plus Samtec’s Sudden Service
philosophy are unique
In ClosingIn Closing • For additional questions in regards to
our
High Speed cable assemblies, please contact our High Speed Cable
Group at:
[email protected]
• For a copy of today’s presentation, please contact us at:
[email protected]
Designing with Samtec High Speed Cable Assemblies
Designing with Samtec High Speed Cable Assemblies