high performance (copper) cable technology
TRANSCRIPT
Jay Diepenbrock
October, 2013
High Performance
(Copper) Cable
Technology
IEEESeptember, 2013 1
Outline
• What and where are “High Performance” cables?
• Cable types
• Differential links
• Cable assembly construction
• Cables and EMI
• Cable EMI mitigation
• Measuring EMC properties of Cables
• References
September, 2013 2
High Performance Cables
September, 2013 3
• Where? Everywhere• What?
• “Big Data” servers, networks• Ethernet, InfiniBand, SAS, PCI-Express
• PCs• SAS, USB 3.0
• Multimedia devices• USB 3.0, Thunderbolt
• TVs, entertainment• Coax (!), HDMI
High Performance Cables
September, 2013 4
40
80
128
1224
6 612
30
60
120
168
300
100.012 0.48 54.95 0
10.218
0.4 4
40
10
100
0
50
100
150
200
250
300
350
1990 1995 2000 2005 2010 2015
Ag
gre
ga
te t
hro
ug
hp
ut,
Gb
/s
Year
I/O Interface Data Rates PCI-Express Gen. 1
PCI-Express Gen. 2
PCI-Express Gen. 3
SAS 2.1
SAS 3
S-ATA 1.0
S-ATA 2.0
S-ATA 3.0
InfiniBand SDR
InfiniBand DDR
InfiniBand QDR
InfiniBand FDR
InfiniBand EDR
Thunderbolt
USB 1.1
USB 2
USB 3.0
HDMI 1.0
HDMI 1.3
HDMI 1.4
HDMI 2.0
Ethernet (100 Mb)
Ethernet (Gb)
Ethernet (802.3ba)
Ethernet (SFF-8431)
Ethernet (802.3bj)
Passive or ActiveCopper or Fiber
Bulk wire construction• Shielded or not• Single or multiconductor + Ground• Round or ribbonized• Flex• Laminated coax• Hybrid – misc. mixes (signals + power, etc.)
Connectors• Coax (F, SMA, N)• Direct attach multi-pin• Paddle card (soldered) multi-pin• Backplane style• “Pluggable” transceiver
Cable types
September, 2013 5
Cable Types
September, 2013 6
connectorconnector
Passive
Cu bulk wire
connectorconnector Cu bulk wire
connectorconnector
Half active(Tx or Rx end)
Cu bulk wire
Full active
connectorconnector Optical fiber
Active Optical
= electrical amp or eq. = O/E or E/O converter
Single-conductor Cable (coax)
September, 2013 7
Construction• Many sizes, materials• Majority are 50 or 75 Ohms• Single signal conductor• Dielectric – PE, PTFE, etc.• Shield (braid or foil+braid)• Jacket
Applications• TV, radio broadcasting• Cable TV• Commercial, amateur radio• Military• Cell phones• Anything RF (audio?)
D
𝑍0 =60
𝑒𝑟
𝐷
𝑑
shield
d
Center
Cond.
dielectric
Differential Pair Cables
September, 2013 8
Majority of high speed interfaces now differential• On chip, between functional islands• Memory• On-card• I/O
Why Differential signaling?• Higher system noise margin
• Power supply voltages decreasing -> lower voltage swing• Lower noise immunity (crosstalk)• Reduced EMI
Differential Pair Bulk wire
September, 2013 9
Construction• Two signal lines, many geometries• Typically 100 Ohms impedance• Twisted or parallel pair• Dielectric – air, PE, PTFE, etc.• Shielded (braid or foil+braid) or not • Jacketed or not
Applications• Networking (“Category”) – UTP, STP• HPC, Supercomputing , I/O (Fibre Channel, PCI-e, SAS, S-ATA,
InfiniBand, Ethernet, etc.) • Computer storage – SAS, S-ATA, USB• Consumer – HDMI, USB, Thunderbolt
Twisted Pair bulk wire
September, 2013 10
Application Lane
speed
# lanes
(pairs)
Cable
Type
Ethernet 1-1000
Mb/s
4 Cat. 3, 5,
5e UTP*
Ethernet 10 Gb/s 4 Cat. 6a
STP
PCI-e 2.5-16
Gb/s
2-32 SPP
FC, Enet,
IB,
2-25 Gb/s 2-24 SPP*
• Inexpensive• Various performance grades• (“Category” 5, 5e, 6, 6a, 7 cables)• Some shielded• Can be field terminated• Susceptible to crosstalk
Shielded Parallel Pair (“Twinax”) bulk wire
September, 2013 11
• Higher performance than TP• Individually Shielded Pairs• Various dielectrics -
PE, PTFE, etc.
• Foil and/or bulk braid shield• Outer jacket per application
• Flammability• Abrasion, chemical resistance
• Applications - I/O, networking
(FC, PCI-e, SAS, S-ATA, InfiniBand, Ethernet, etc.)
Bulk shield
dielectric
s
Drain wired
D
Shielded Parallel Pair (“Twinax”)
September, 2013 12
Advantages• Good performance• Low crosstalk
Pitfalls• Symmetry important
• Non-uniform materials• Geometric structure
• Common Mode generation• Skew• System asymmetries
• Manufacturing
good
bad
Shielded Parallel Pair shield topology
September, 2013 13
EXD versus Standard Spiral Shield 24 AWG 100 Ohm
-100
-90
-80
-70
-60
-50
-40
-30
-20
-10
0
0 2000 4000 6000 8000 10000 12000 14000 16000 18000 20000
Frequency MHz
SD
D21 d
B / 1
0 m
ete
r
thru fixture
EXD 1
EXD 2
EXD 3
EXD 4
Optimized for High Frequency 1
Optimized for High Frequency 2
Optimized for High Frequency 3
Optimized for High Frequency 4
EXD
Spiral 10 meter data, fixture not removed
Longitudinal
shield
Spiral shield
Quad
September, 2013 14
Construction• Four signal lines (two pairs), but smaller• Dielectric – PE, PTFE• Unshielded quad• Bulk shield• Jacket
Applications• HPC, Supercomputing • Limited usage
• Expensive, hard to make – orthogo-nality critical to CM, xtalk perf.
• Hard to terminate
1+
1-
2+ 2-
shield
Connectors
September, 2013 15
7-pin Serial ATA right-angle 7-pin Serial ATA straightSFF-8482 SAS 29 pin w /power
SFF-8470 SAS SFF-8088 external mini SAS SFF-8487 internal mini SAS
HDMI
Connectors
September, 2013 16
PCI-Expressx16
SFF-8038 (SFP+)
(QSFP)
September, 2013 17
Raw Cable
Spring
Cover Shell
Screw
PCBA
Latch
Base
Insert Molding
Spacer
Cover
Tear-down – QSFP (SFF-8088)
Wire termination
September, 2013 18
QSFPSFF-8088 (12X InfiniBand)
Differential Links
Each signal transmitted by a pair of conductors, driven
180 degrees out of phase
Considerations:
–greater common mode noise immunity than single-ended
–less EMI radiation than single-ended
–must consider and measure differential quantities
analysis, simulation methods
test equipment, fixtures
–additional propagation modes are possible
+ -
Signal conductors
Drain wire Foil shield
Dielectric
+ -
card wirecable
September, 2013 19
Differential Impedance
• “Modes" are now possible
• Case 1
L/CC11 C12
C21 C22
L11 L12
L21 L22
L, C, Z
L, C, Zcommon mode
September, 2013 20
Differential Impedance
• “Modes" are now possible
• Case 1
L/CC11 C12
C21 C22
L11 L12
L21 L22
L, C, Z
L, C, Z
common mode
• Case 2
L, C, Z
L, C, Z
differential mode
the modes have different impedances,
and different propagation delays!
It's still , but now C= and L=
September, 2013 21
Differential Measurements
• Options–Make multiple single-ended measurements and do the math yourself
–Buy differential test equipment, build differential fixtures
Differential TDR - measure M1=C1-C2
Four port VNA or two port with external test set - measure sdd21, not s21,
and sdd11, not s11
Provides additional information over use of baluns (no common mode data)
Z11 Z12 Z22
see Carey, Scott, and Weeks: "Characterization of Multiple Parallel Transmission Lines,"
IEEE Trans. Instr. and Meas., Sept. 1969
September, 2013 22
Differential Pair Skew
• Two types:–in-pair (between legs of pair)Due to difference in propagation delay between legs of pair
Manifested as "excess attenuation"
Spec. limits pretty tight - causes differential imbalance, and can
cause EMI problems due to common mode energy
not uniform with length!
–pair to pair (between pairs)difference in propagation delay between pairs
modern interfaces relatively insensitive to it (500 ps limit) - it's
corrected in the design
September, 2013 23
Skew
September, 2013 24
Skew
September, 2013 25
• Small amounts of skew create significant common mode noise
• As little as 1% of bit width for skew can have significant EMI effects
• As little as 10% of bit width skew creates CM signal of equivalent amplitude to initial signals
Skew
September, 2013 26
Individual Channels of Differential Signal with Skew
2 Gb/s with 50 ps Rise and Fall Time (+/- 1.0 volts)
-0.6
-0.4
-0.2
0
0.2
0.4
0.6
5.0E-10 1.0E-09 1.5E-09 2.0E-09 2.5E-09 3.0E-09
Time (seconds)
Vo
lta
ge
Channel 1
No Skew
10 ps
20 ps
50 ps
100 ps
150 ps
200 ps
Skew
September, 2013 27
Common Mode Voltage on Differential Pair Due to In-Pair Skew
2 Gb/s with 50 ps Rise and Fall Time (+/- 1.0 volts)
-0.6
-0.4
-0.2
0.0
0.2
0.4
0.6
5.0E-10 1.0E-09 1.5E-09 2.0E-09 2.5E-09 3.0E-09 3.5E-09 4.0E-09 4.5E-09 5.0E-09
Time (seconds)
Am
pli
tud
e (
vo
lts)
10 ps
20 ps
50 ps
100 ps
150 ps
200 ps
Rise/fall time mismatch
September, 2013 28
• Small amounts of mismatch create significant CM noise
• Not as significant as skew, but harder to control!• Telltale is significant 2nd harmonic content
Rise/fall time mismatch
September, 2013 29
Example of Effect for Differential Signal with Rise/Fall Time Mismatch
2 Gb/s Square Wave (Rise/Fall = 50 & 100 ps)
-0.6
-0.4
-0.2
0
0.2
0.4
0.6
0.0E+00 2.0E-10 4.0E-10 6.0E-10 8.0E-10 1.0E-09 1.2E-09 1.4E-09 1.6E-09 1.8E-09 2.0E-09
Time (Seconds)
Vo
ltag
e
Channel 1
Channel 2
T/R=50/100ps
Rise/fall time mismatch
September, 2013 30
Common Mode Voltage on Differential Pair Due to Rise/Fall Time Mismatch
2 Gb/s with Differential Signal +/- 1.0 Volts
-0.2
-0.15
-0.1
-0.05
0
0.05
0.1
0.15
0.2
0 5E-10 1E-09 1.5E-09 2E-09 2.5E-09 3E-09 3.5E-09 4E-09 4.5E-09 5E-09
Time (seconds)
Level
(vo
lts)
T/R=50/100ps
T/R=50/150ps
T/R=50/200ps
Eye opening and Jitter
Measures time domain performance of link
Measured using PRBS or application-specific data pattern (e. g., CJTPAT)
Eye opening -
–vertical "black space" in middle of many overlaid bits
–minimum opening needed for receiver to distinguish between "1" and "0"
Jitter - horizontal width of zero crossing of overlaid waveforms
eye opening
jitter
September, 2013 31
Eye Opening and Jitter – test setup
September, 2013 32
Asynch. Crosstalk Source
Test
card
PRBS7, 9, ..31 pattern
Vout ~= 1 Vpp
Trise ~= 30 ps
xx Gb/s
Color-graded display
Infinite persistence
x Histogram hits
(terminate unused ports
with 50 Ohms to Ground)
Clock
Test
card
Cable
Pattern or BERT Gen. Sampling or real-time oscilloscope
Sources of EMI in Cables
September, 2013 33
• Skew in system coupled to cable shield, due to• Asymmetric differential pairs• Unequal rise/fall time of signals• Common mode in signals
• Cable construction• Common mode conversion in bulk wire• Poor connection from Chassis to Cable plug backshell• Leaky backshell• Skew in plug/paddle card/bulk wire• Poorly shielded bulk wire
Common mode conversion
September, 2013 34
Cable Assembly Construction Influence on EMC
September, 2013 35
• Shielding• Pair shields – foil in or out? Shielded or not? Drain wire handling• Bulk shield
• Foil (high freq.)• Braid (low freq.) – shield coverage (typ. 80-90%), weave angle, etc.
• Backshell design• Seams, leakage potential• Latches, jack screws
• Grounding
• Backshell-chassis connection – springs, gaskets, drain wires
• Don’t forget the system influence!• In-pair skew• Mismatched rise/fall tmes• Common mode
Cable EMI sources
September, 2013 36
From Bergey and Altland, “EMI Shielding of Cable Assemblies”, DesignCon 2008
HDMI cable shield connection
Cable EMI sources
September, 2013 37
USB cable shield connection (or not!)
From Bergey and Altland, “EMI Shielding of Cable Assemblies”, DesignCon 2008
Cable EMI sources
September, 2013 38
From Bergey and Altland, “EMI Shielding of Cable Assemblies”, DesignCon 2008
Measuring Cable EMI
September, 2013 39
• Key parameters• Transfer Impedance• Shielding Effectiveness
• Measurement methods• EM 52022 (CISPR 22) – semi-anechoic chamber• Tube fixture (IEC 62153-4-7)
• Measures transfer impedance• Max. frequency ~1 GHz
• Reverb chamber (no standard yet)• Measures shielding effectiveness• Usable ~300 MHz – 20 GHz
Tube Fixture
September, 2013 40
Tube Fixture Sample Results
September, 2013 41
Reverb Chamber
September, 2013 42
• Closed, conductive-walled room• Usable frequency range ~300 MHz-20 GHz, depending on
room size and antennae used• Don’t dampen resonance, celebrate it!• CUT is driven with differential or common mode signal,
radiated energy is measured• No system hardware required• “Tuner” used to stir resonances, either stepped or
continuously from external controller• Much work on reverb chambers at OK State Univ.
(C. Bunting, et. al.)
September, 2013 43
Reverb Chamber
Reverb Chamber
September, 2013 44
MeasurementAntennas
CUT support (non-conductive)
CUT
Tuner
Stepper motor
September, 2013 45
• Many paths to EMC cleanliness• Reduce system in-pair skew• Match signal rise/fall times• Reduce common mode energy coupling to cable shield• Improve cable shield connection to cable backshell
• Reduce connection inductance• Better shield coverage
• Utilize absorbing material in cable jacket• Utilize Band Gap devices on host card
EMI Mitigation in Cables
September, 2013 46
EMI Absorbing Material
September, 2013 47
• Available from ARC Technologies, Inc. for• extrusion in cable jacket• Molded enclosures (replace metal can)• Covers over connectors
• Frequency selective – suppression range depends on formula used• Doesn’t need to be used on whole cable – just ends are enough
EMI Absorbing Material
September, 2013 48
Motivation - Eliminate Ferrite Cores on Cables
EMI Absorbing Material
September, 2013 49
Ethernet Cable Emission Reduction (When Drive Signal at Same End of Cable)
ARC Lossy Material Covers Partial Length
0
2
4
6
8
10
12
14
16
18
20
0.0E+00 1.0E+09 2.0E+09 3.0E+09 4.0E+09 5.0E+09 6.0E+09 7.0E+09 8.0E+09 9.0E+09 1.0E+10
Frequency (Hz)
Re
du
cti
on
in
Em
issio
ns
(d
B)
Ethernet Sample #1 w/ 11" Covered
Ethernet Sample #1 w/ 23" Covered
Ethernet Sample #1 w/ 37" Covered
Ethernet Sample #1 Full Cable Covered
References
September, 2013 50
• Diepenbrock, J.: Measurement and Analysis of Shielding Effectiveness and Transfer Impedance of High Speed Data Cables, DesignCon 2012
• Archambeault, B., Connor, S., Diepenbrock, J., and Knight, A.: Developing Limits for Common Mode Noise on High Speed Differential Signals, DesignCon 2011
• Hill. D.: “Electromagnetic Theory of Reverberation Chambers,” Natl. Inst. of Standards and Technology Tech Note 1506, 1998
• Vignesh Rajamani, Charles F. Bunting and James C. West, “Calibration of a Numerically Modeled Reverberation Chamber,” IEEE Symposium on Electromagnetic Compatibility 2009
• Archambeault, B., Chikando, E., Connor, S., and Diepenbrock, J.: “High SpeedCables with Lossy Material Coating,” IEEE 2010 Symposium on Electromagnetic Compatibility 2010
Other ReferencesStandards• Code of Federal Regulations Title 47, Telecommunications, part 15 (US)• EN 55022, Information Technology Equipment – Radio Disturbance Characteristics – Limits and
Methods of Measurement (Europe)• ANSI/EIA/ECA 364-66A EMI Shielding Effectiveness of Electrical Connectors• IEC 61000-4-21 Reverb chamber test methods• IEC 61276 Screening attenuation measurement by the reverberation chamber method• IEC 62153-4-7 Transfer impedance and screening, tube in tube method• IEC 62153-4-9 Coupling attenuation of screened balanced cables, triaxial method• IEEE 802• InfiniBand Specification, volume 2• PCI-Express Cabling Specification
Other• Agilent Technologies: Understanding the Fundamental Principles of Vector Network
Analysis," AN 1287-1, available at http://www.agilent.com
• Bogatin, E: "Differential Impedance Finally Made Simple,“ available at
http://www.ewh.ieee.org/r5/denver/rockymountainemc/archive/2000/diffimp.pdf
• Carey, Scott, and Weeks: "Characterization of Multiple Parallel Transmission Lines,"
IEEE Trans. Instr. and Meas., Sept. 1969
• Deutsch, A., "Electrical Characteristics of Interconnections for High-Performance
Systems," IEEE Proceedings vol. 86 No. 2, Feb. 1998
September, 2013 51
IEEE10/30/2013 52
Conferences
• DesignCon – February, in Santa Clara, CA• IEEE Electrical Performance of Electronic Packaging (EPEP)• IEEE EMC Symposium (EMCS)
• in Raleigh, NC in August, 2014• Embedded SI conference• http://www.emcs.org
• IEEE ECTC, ED, ISSCC• IEEE SPI workshop (Europe)