Technical Seminar

Document Sample
Technical Seminar Powered By Docstoc
					Technical Seminar - H316

   Evaluation of Maximum Usable
    Lengths for Cabled Copper
           Interconnects

  H316: High-performance System
        Design Conference
                 Michael W. Fogg
                                        Ltd.
  AMP Incorporated/A tyco International Ltd. Company
             mike.fogg@ tycoelectronics.com
             mike.fogg@ tycoelectronics.com


                            1
Outline
• Current Technology/Emerging
  Technologies
• 100 vs 150 Ohm Question
• 1X/4X/12X Proposals
• Conductor Losses
• Equalization Approaches
• Connector Issues
• Calculations
• Conclusions
                  2
Outline
• Current Technology/Emerging
  Technologies
  – Fibre Channel
  – Gigabit Ethernet
  – 10Gig Ethernet
  – HARI
  – Infiniband (IB)




                       3
Outline
• 100 vs 150 Ohm Question
  – Current Product is 150 Ohm
  – Proposals are 100 Ohm
  – PCB Construction
  – Cable Construction
    • Dielectrics
    • Size/AWG Tradeoffs




                      4
100 vs 150 Ohm Question
• Current Product is 150 Ohm
  – Fibre Channel
  – Gigabit Ethernet (802.3z)
• Proposals are 100 Ohm
  – HARI (4X)
  – Parallel (12X)
  – Serial >2.5 Gbps ?




                      5
100 vs 150 Ohm Question
• PCB Construction
                                        – Problem is worse for
                                          differential
                                                                             Board Construction for 1/2 oz Copper
                                                         0.05
  Reference plane to reference plane spacing (inches)




                                                        0.045          0.005" Trace
                                                                       0.008" Trace
                                                         0.04

                                                        0.035

                                                         0.03

                                                        0.025

                                                         0.02

                                                        0.015

                                                         0.01

                                                        0.005
                                                             35   40    45       50       55      60      65        70   75   80
                                                                               Characteristic Impedance, Ohms



                                                                                                                6
Outline
• 1X/4X/12X Proposals
  – Serial Approaches (1X)
     • 1 Pair Each Direction
  – ‘Striped’ Approaches (4X)
     • HARI
     • 4 Pairs Each Direction
  – Parallel Approaches (12X)
     • 8 to 12 Pairs Each Direction
  – Evaluation of Maximum AWG
     • “Fat Pipe” Calculations




                          7
1X/4X/12X Proposals
• Serial Approaches (1X)
  – 1 Pair Each Direction
  – Similar to existing Fibre Channel and
    Gigabit Ethernet (802.3z)
  – Concern about implementation of 10-
    12.5Gbps data rates




                      8
1X/4X/12X Proposals
• ‘Striped’ Approaches (4X)
  – HARI
  – 4 Pairs Each Direction
  – Increase signal count to simplify
    implementation
     • 2.5 to 3.125Gbps X 4 pairs each direction
       instead of 10 to 12.5Gbps X 1 pair each way
  – Requires larger connector (maybe)




                         9
1X/4X/12X Proposals
• Parallel Approaches (12X)
  – 8 to 12 Pairs Each Direction
  – True Parallel
  – Skew Concerns
  – More data pairs requires reduction in
    conductor size
  – Requires larger connector




                      10
1X/4X/12X Proposals
• ‘Fat Pipe’ Calculations
  – Assumptions:
     •   Fixed maximum diameter
     •   Based on existing 22AWG, 150 Ohm Quad
     •   Diameter over dielectric 0.267”/6.78mm
     •   3 Typical Dielectrics
          – Polyethylene (PE)
          – Foamed Polyethylene (F-PE)
          – Expanded PTFE (E-PTFE)
     • Use Industry Standard ‘K’ Factors




                           11
1X/4X/12X Proposals
• ‘Fat Pipe’ Calculations
  – Assumptions:
     • Outer diameter is constant, wire AWG changes

                        Serial/Quad Cable           Serial/2 Pair Cable
                                         Signal 2+
               Foil Shield
         Braid Shield




                                             Signal 1-
             Signal 1+
                Insulation
                                        Signal 2-



                             4X/Quad Cable               4X/8 Pair Cable




                                         12
1X/4X/12X Proposals
• ‘Fat Pipe’ Calculations
  – 150 Ohm Constructions
      • 22 AWG Quad used for comparisons

   Assembly   Cable       Maximum AWG (150 Ohm)
   Type       Description   PE     F-PE E-PTFE
   1X-True    2 Pair        28      26      24
   Serial     1 Quad        22      20      19
                                       *
   4X-HARI    8 Pair        34      32      31
              4 Quad        30      28      27
                                       *
   12X-       24 Pair       39      37      36
   Parallel   12 Quad       34      32*     31

                      *Properties of foamed polyethylene
                      make this difficult to manufacture

                         13
1X/4X/12X Proposals
• ‘Fat Pipe’ Calculations
  – 100 Ohm Constructions
      • 26 AWG pair used for 4X comparisons
      • 30 AWG pair used for 12X comparisons
   Assembly   Cable       Maximum AWG (100 Ohm)
   Type       Description   PE     F-PE E-PTFE
   1X-True    2 Pair        24      23      22
   Serial     1 Quad        17      16      15
   4X-HARI    8 Pair        29      27      26
              4 Quad        25      24      23
                                      *
   12X-       24 Pair       33      31      30
   Parallel   12 Quad       29      28      27
                       *Properties of foamed polyethylene
                       make this difficult to manufacture

                          14
Outline
• Conductor Losses
  – Sources of Loss
    •   DC Loss
    •                      (sqrt(f))
        Skin Effect Losses (sqrt(f))
    •   Dielectric Losses
    •   Reflective Losses
    •   Leakage (EMI)
  – Comparison of Cable vs. PCB
    • Comparison of Conductor Loss
    • Comparison of Dielectric Loss




                            15
Conductor Loss
• Conductor Losses
  – Sources of Loss
    •   DC Loss
    •                      (sqrt(f))
        Skin Effect Losses (sqrt(f))
    •   Dielectric Losses
    •   Reflective Losses
    •   Leakage




                            16
Conductor Loss
              – Comparison of Cable vs. PCB Losses
                                     dB loss/m, Cables and PCB
              0
                                                                       22 AWG, 150 Ohm
              -1
                                                                                             Comparative Loss,
                                                        26 AWG, 100 Ohm
                                                                                             dB/m
              -2                                                                             1. 22 AWG, 150 Ohm
              -3
                                                                                             Cable
                                               Rogers 4350                                   2. 26 AWG, 100 Ohm
 Loss, dB/m




              -4
                                     FR-4
                                                                                             Cable
              -5
                                                                                             3. PCB trace, .012”
                                                                                             Rogers 4350, 50 Ohm
              -6                                                                             4. PCB trace, .012”,
              -7
                                                                                             FR-4, 50 Ohm

              -8

              -9
                   0   0.2   0.4   0.6      0.8     1     1.2        1.4   1.6   1.8     2
                                             Frequency, GHz




                                                                17
Conductor Loss
      – Comparison of 100 Ohm vs 150 Ohm Cable
                         Loss Comparison Between Similar Sized 100 and 150 Ohm Cables
                0
                                                                                                  Comparative Loss,
                                                                  100 Ohm, 22AWG Quad
              -0.2
                                                                  150 Ohm, 28AWG Quad             dB/m
                                                                                                  1. 28 AWG, 150 Ohm
              -0.4                                                                                Cable
                                                                                                  2. 22 AWG, 100 Ohm
              -0.6
                                                                                                  Cable
 Loss, dB/m




              -0.8
                                                                                                  -Both have same
                                                                                                  dielectric diameter
                -1


              -1.2


              -1.4


              -1.6
                     0    0.2    0.4    0.6    0.8    1     1.2     1.4   1.6    1.8          2
                                                Frequency, Hz                                 9
                                                                                       x 10




                                                                   18
Outline
• Equalization Approaches
  – Passive Equalization
  – Pre-emphasis
  – Active Equalization
     • Pre-set
     • Learning Sequence




                       19
Equalization Approaches
• Passive Equalization
  – Common and inexpensive
  – Frequently built into assembly
  – Requires knowledge of data rate and loss
    characteristics of assembly
  – Does not provide gain




    Equalized Output        Unequalized Output


                       20
Equalization Approaches
• Pre-emphasis
  – Requires advance knowledge of loss
    characteristics of interconnect
  – Pre-distortion is ‘corrected’ by cable/PCB
    loss
  – Difficult to change for changes in
    transmission media
  – Silicon/power issues




                       21
Equalization Approaches
• Active Equalization
  – Pre-set
     • Requires advance knowledge of loss
       characteristics of interconnect
     • Selective gain corrects distortion from
       cable/PCB
     • Silicon/Power Issues
  – Learning Sequence
     • Complexity issues
     • Selective gain corrects distortion from
       cable/PCB
     • Silicon/Power Issues



                          22
Outline
• Connector Issues
  – Reflective Loss
  – Crosstalk
  – EMI/RFI
  – Conductor Size/Contact Spacing
  – Example-HSSDC II




                     23
Connector Issues
• The application of a connector will
  degrade the transmitted signal
• Problem gets worse at higher data rates        Transmitted Eye Patterns for 10ft Cable and 10ft Cable Assembly
                                       900

                                                                            100 Ohm, 26AWG Shielded Pair Cable
                                       800
                                                                            100 Ohm, 26 AWG Assembly

                                       700




               Eye Opening at 50% UI
                                       600

                                       500

                                       400

                                       300

                                       200

                                       100

                                         0
                                             0        2         4          6         8         10        12        14
                                                                      Data Rate, Gigabits/s



                                                     24
Outline
• Calculations
  – Proof of Modeling
  – Serial (1X) Links
  – HARI (4X) Links
  – Parallel (12X) Links
  – Comparison of Aggregate Data Rate




                    25
Calculations
• Comparison of Predicted vs. Test
  – 22 AWG, 150 Ohm Polyethylene Quad
  – 17m, 2.5Gbps




                                      Predicted: Duration-0.95 UI
   Test: Duration-0.91 UI
                                      Amplitude-0.40 (normalized)
   Amplitude-0.38 (normalized)




                                 26
Calculations
• Comparison of Predicted vs. Test
  – 22 AWG, 150 Ohm Polyethylene Quad
                                       Tested vs. Predicted Eye Opening, 9m, 22AWG Cable
                              1

                             0.9                                         Tested Eye Opening
                                                                         P redicted Eye Opening

                             0.8

                             0.7
    Normalized Eye Opening




                             0.6

                             0.5

                             0.4

                             0.3

                             0.2

                             0.1

                              0
                                   0    1          2          3           4           5           6
                                                       Frequency, Gbps


                                                               27
Calculations
• Comparison of Maximum Lengths
  – Serial (1X) Approaches
     • 22 AWG, 150 Ohm Polyethylene Quad
     • 10Gbps/12.5Gbps
  – HARI (4X) Approaches
     • 26 AWG, 100 Ohm E-PTFE Pairs
     • 2.5Gbps/3.125Gbps
  – Parallel (12X) Approaches
     • 30 AWG, 100 Ohm E-PTFE Pairs
     • 1.0Gbps/1.25Gbps
  – 1000mV Launch, 200mV Receive
  – With/Without Passive Equalization

                       28
Calculations
• Comparison of Maximum Lengths
  – Serial (1X) Approach
  – 22 AWG, 150 Ohm Polyethylene Quad
                                    Predicted Performance, 22 AWG, 150 Ohm Serial
                       50
                                                                       Unequalized 22AWG
                       45                                              Equalized 22AWG

                       40

                       35

                       30
      Length, meters




                       25

                       20

                       15

                       10

                        5

                        0
                            0   2          4         6          8      10           12     14
                                                Data Rate/line, Gbps



                                                          29
Calculations
• Comparison of Maximum Lengths
  – HARI (4X) Approach
  – 26 AWG, 100 Ohm E-PTFE Pair
                                    P redicted Performance, 26 AWG, 100 Ohm 4X
                     15




                     10
    Length, meters




                      5



                                    Unequalized 26AWG
                                    Equalized 26AWG


                      0
                          0   0.5    1    1.5     2     2.5      3     3.5   4   4.5   5
                                                Data Rate/line, Gbps



                                                                30
Calculations
• Comparison of Maximum Lengths
  – Parallel (12X) Approach
  – 30 AWG, 100 Ohm E-PTFE Pair
                                      Predicted Performance, 30 AWG, 100 Ohm 12X
                       15

                                                                         Unequalized 30AWG
                                                                         Equalized 30AWG




                       10
      Length, meters




                        5




                        0
                            0   0.5   1     1.5     2      2.5     3     3.5    4    4.5     5
                                                  Data Rate/line, Gbps



                                                            31
Calculations
• Comparison of Maximum Lengths
  – Total Aggregate Data (1X/4X/12X)
                                 P r e d i c t e d P e r f o r m a n c e , 1 2 . 5 G b p s A g g r e g a t e D a t a , 1 X/4X/12X
                        18


                        16


                        14


                        12
       Length, meters




                        10


                         8


                         6                  Unequalized Serial
                                            Equalized Serial
                                            Unequalized 4X
                         4
                                            Equalized 4X
                                            Unequalized Parallel
                         2                  Equalized Parallel


                         0
                             8               9             10       11       12                           13             14         15
                                                           Data Rate/assembly, Gbps



                                                                          32
Outline
• Conclusions
  – How Fast/How Far
  – Practical Considerations
     •   Cost of Cable
     •   Cost of Termination
     •   Size of Connector
     •   EMI/RFI
     •   Silicon
  – What’s Next?
     • How to go farther




                           33
Conclusions
• How Fast/How Far
  – Parallel approaches provide longest length
    at 12.5 Gbps (13m vs 10m) with passive
    equalization
  – Parallel approaches provide longest length
    at 12.5 Gbps (10m vs 8m) without
    equalization
  – The addition of connectors will reduce
    lengths




                      34
Conclusions
• Practical Considerations
  – Cost of Cable
     • More conductors=higher cost
  – Cost of Termination
     • More conductors=higher cost
  – Size of Connector
     • More conductors=larger connector
     • Conductor size for 100 Ohm serial solutions
  – EMI/RFI
                        vs.
     • Higher Frequency vs. More Power
  – Silicon
     • SERDES
     • Equalization/Pre-emphasis

                         35
Conclusions
• What’s Next?
  – How to go farther
     • Fill in the Matrix
          – Finish the 100 vs 150 ohm comparison study
     •   Investigation of Pre-emphasis
     •   Investigation of Active Equalization
     •   Different Modulation Schemes
     •   Fiber Optics




                             36

				
DOCUMENT INFO