lecture7 ee689 eq intro txeq

Upload: dogudogu

Post on 02-Jun-2018

221 views

Category:

Documents


0 download

TRANSCRIPT

  • 8/10/2019 Lecture7 Ee689 Eq Intro Txeq

    1/27

    Sam Palermo Analog & Mixed-Signal Center

    Texas A&M University

    ECEN689: Special Topics in High-SpeedLinks Circuits and Systems

    Spring 2012

    Lecture 7: Equalization Introduction & TX FIR Eq

  • 8/10/2019 Lecture7 Ee689 Eq Intro Txeq

    2/27

    Announcements

    Exam 1 is March 7 5:45-7:10PM (10 extra minutes) Closed book w/ one standard note sheet

    8.5x11 front & back Bring your calculator Covers material through lecture 6

    Previous years exam 1s are posted on thewebsite for reference

    2

  • 8/10/2019 Lecture7 Ee689 Eq Intro Txeq

    3/27

    Agenda

    Equalization theory and circuits Equalization overview Equalization implementations

    TX FIR RX FIR RX CTLE RX DFE

    TX FIR Equalization FIR filter in time and frequency domain MMSE Coefficient Selection Circuit Topologies

    Equalization overview paper posted on website 3

  • 8/10/2019 Lecture7 Ee689 Eq Intro Txeq

    4/27

  • 8/10/2019 Lecture7 Ee689 Eq Intro Txeq

    5/27

  • 8/10/2019 Lecture7 Ee689 Eq Intro Txeq

    6/27

    6

    Channel Performance Impact

  • 8/10/2019 Lecture7 Ee689 Eq Intro Txeq

    7/27

  • 8/10/2019 Lecture7 Ee689 Eq Intro Txeq

    8/27

    Channel Equalization

    8

    Equalization goal is to flatten the frequency response out to theNyquist Frequency and remove time-domain ISI

  • 8/10/2019 Lecture7 Ee689 Eq Intro Txeq

    9/27

    TX FIR Equalization

    TX FIR filter pre-distorts transmitted pulse inorder to invert channel distortion at the cost ofattenuated transmit signal (de-emphasis)

    9

    L

    L L

    L

    L

    L

    L

    L

    L

    1x 4x 2x 1x

    1/4 1 1/2 1/4IDACs&

    BiasControl

    sgn -1 sgn 0 sgn 1 sgn 2

    50

    Out-P

    Out-N

    4:2MUX

    2

    2

    2

    21

    D0

    D1

    D2

    D3

    VDDA=1.2VVDD=1.0V

    VDDIO=1.0V

    VDDA=1.2V

    1

    1

    1

    C2 (5GHz)

    From on-chip PLL

    2

    ( 2 . 5

    G b / s )

    (10Gb/s)

    (5Gb/s)

    ESD

    L

    L L

    L

    L

    L

    L

    L

    L

    LL

    LL LL

    LL

    LL

    LL

    LL

    LL

    LL

    1x 4x 2x 1x

    1/4 1 1/2 1/4IDACs&

    BiasControl

    sgn -1 sgn 0 sgn 1 sgn 2

    50

    Out-P

    Out-N

    4:2MUX

    2

    2

    2

    21

    D0

    D1

    D2

    D3

    VDDA=1.2VVDD=1.0V

    VDDIO=1.0V

    VDDA=1.2V

    1

    1

    1

    C2 (5GHz)

    From on-chip PLL

    2

    ( 2 . 5

    G b / s )

    (10Gb/s)

    (5Gb/s)

    ESD

    ] 2

    210102101

    TERM out

    R D I D I D I D I V

    A Low Power 10Gb/s Serial Link Transmitter in 90-nmCMOS, A. Rylyakov et al., CSICS 2005

    I -1 I 0 I 1 I 2

    D(1) D(0) D(-1) D(-2)

  • 8/10/2019 Lecture7 Ee689 Eq Intro Txeq

    10/27

  • 8/10/2019 Lecture7 Ee689 Eq Intro Txeq

    11/27

    RX Equalization #1: RX FIR

    Pros With sufficient dynamic range, can amplify

    high frequency content (rather thanattenuate low frequencies)

    Can cancel ISI in pre-cursor and beyondfilter span

    Filter tap coefficients can be adaptivelytuned without any back-channel

    Cons Amplifies noise/crosstalk Implementation of analog delays Tap precision

    11

    w -1

    z -1

    x w 0

    z -1

    x

    z -1

    x w n-1

    z -1

    w nx

    DEQ

    D in

    Analog Delay Elements

    *

    *D. Hernandez-Garduno and J. Silva-Martinez, A CMOS 1Gb/s 5-Tap Transversal Equalizer based on 3 rd-Order Delay Cells,"ISSCC, 2007.

  • 8/10/2019 Lecture7 Ee689 Eq Intro Txeq

    12/27

    RX Equalization #2: RX CTLE

    12

    Din- D in+

    Vo-Vo+

    Pros Provides gain and

    equalization with lowpower and areaoverhead

    Can cancel both pre-cursor and long-tail ISI

    Cons Generally limited to 1st

    order compensation Amplifies noise/crosstalk PVT sensitivity

    Can be hard to tune

  • 8/10/2019 Lecture7 Ee689 Eq Intro Txeq

    13/27

    RX Equalization #3: RX DFE

    13

    z -1clk

    x

    w 1

    z -1x

    w 2

    z -1x

    w n-1

    z -1x

    w n

    D in D RX

    Pros No noise and crosstalk

    amplification Filter tap coefficients

    can be adaptively tunedwithout any back-channel

    Cons Cannot cancel pre-

    cursor ISI Critical feedback timing

    path Timing of ISI

    subtraction complicatesCDR phase detection

  • 8/10/2019 Lecture7 Ee689 Eq Intro Txeq

    14/27

    Equalization Effectiveness

    Some observations: Big initial performance boost with 2-tap TX eq. With only TX eq., not much difference between 2 to 4-tap RX equalization, particularly DFE, allows for further performance

    improvement Caution hard to build fast DFEs due to critical timing path 14

    I n c r e a s

    i n g

    E q u a

    l i z a t

    i o n

    Channel Responses

  • 8/10/2019 Lecture7 Ee689 Eq Intro Txeq

    15/27

    Link with Equalization

    15

    S e r i a l i z e r

    DTX[N:0]

    TX ClkGeneration

    (PLL)

    TX FIREqualization

    RX ClkRecovery

    (CDR/Fwd Clk)

    RX CTLE + DFEEqualization

    D e s e r i a l i z e r

    DRX[N:0]

    Channel

    f

  • 8/10/2019 Lecture7 Ee689 Eq Intro Txeq

    16/27

    Channel Equalization

    16

    Equalization goal is to flatten the frequency response out to theNyquist Frequency and remove time-domain ISI

  • 8/10/2019 Lecture7 Ee689 Eq Intro Txeq

    17/27

  • 8/10/2019 Lecture7 Ee689 Eq Intro Txeq

    18/27

    TX FIR Equalization Freq. Domain

    18

    z -1

    z -1

    z -1

    w -1

    w 0

    w 1

    w 2

    TX

    data

    z -1 w n

    ( ) 21 274.0595.0131.0 += z z zW :10GbpsFor

    ( )( )ss fT j fT j fT e z

    z z zW s

    2sin)2cos(

    274.0595.0131.02

    21

    +==+=

    w/

    ( )( ) ( ) dB f W j z

    f

    4.14190.001)0sin(0cos ===+== 0ResponseFrequencyLow

    ( ) dBT

    f W j z

    T f

    s

    s

    0121

    1)sin(cos

    21

    =

    ==+=

    =

    ResponseFrequencyNyquist

    Equalizer has 14.4dB of frequency peaking

    Attenuates DC at -14.4dB and passes Nyquist frequency at 0dB

    Note: T s=T b=100ps

  • 8/10/2019 Lecture7 Ee689 Eq Intro Txeq

    19/27

  • 8/10/2019 Lecture7 Ee689 Eq Intro Txeq

    20/27

    TX FIR Coefficient Selection

    Multiplying input symbols by TX Eq., wc=w*c

    20

    ( )( )

    ( )

    ( )( ) ( )

    ( ) ( )( )

    ( )( ) ( )

    ( ) ( )( )

    ( )( )

    ( )

    =

    ++ 1...

    1

    0

    10...000

    21...000

    ..................

    00...001

    00...000

    10...000

    21...000

    ..................

    00...001

    00...000

    3

    ...

    1

    0

    lc

    c

    c

    nw

    nwnw

    ww

    w

    k h

    k hk h

    hh

    h

    k nl y

    y

    y Total system

    ( )( )

    ( )

    ( )( ) ( )

    ( ) ( )( )

    ( )( )

    ( )+

    =

    ++ 1...

    1

    0

    10...00021...000

    ..................

    00...001

    00...000

    3

    ...

    1

    0

    lnwc

    wc

    wc

    k hk hk h

    hh

    h

    k nl y

    y

    y

    We desire the output vector, y, to be ISI free( )( ) ++=

    ++===

    1#tapprecursorEq#samplecursor-preChannel

    1#tapprecursorEq#samplecursor-preChannel

    nn y

    nn y y

    des

    desdes ,0

    ,1

  • 8/10/2019 Lecture7 Ee689 Eq Intro Txeq

    21/27

    Lone-Pulse Equalization Example

    21

    With lone-pulse equalization,l =1 input symbols, i.e. c=[1]

    ( )

    ( )

    ( )

    [ ]1

    2

    1

    0

    0.0067 0 0

    0.0090 0.0067 0

    0.0097 0.0090 0.0067

    0.0152 0.0097 0.0090

    0.0162 0.0152 0.0097

    0.0224 0.0162 0.0152

    0.0360 0.0224 0.01620.0526 0.0360 0.0224

    0.0917 0.0526 0.0360

    0.1775 0.0917 0.0526

    0.3437 0.1775 0.0917

    0.0812 0.3437 0.1775

    0.0052 0.0812 0.3437

    0.0023 0.0052 0.0812

    0.0010 0.0023 0.0052

    0.0004 0.0010 0.0023

    0 0.0004 0.0010

    0 0 0.0004

    0

    0

    0

    0

    0

    0

    00

    0

    0

    0

    1

    0

    0

    0

    0

    0

    0

    =

    w

    w

    w

    Y des

    Channel pulse matrix H with 5 pre-

    cursor samples and 10 post-cursorsamples, 3 columns for 3 eq taps

    3-tap EqMatrix, W

    Symbol Matrix,C for

    Lone Pulse

    Y des (5+1+1=7)=1

    Channel pre-cursor samples

    Equalization pre-cursor taps

  • 8/10/2019 Lecture7 Ee689 Eq Intro Txeq

    22/27

    TX FIR Coefficient Selection

    Differentiating this w.r.t. tap matrix taps to find taps which yield

    minimum error norm2

    22

    inputpulsewithdesdesC des Y HW Y HW Y Y E === We can calculate the error w.r.t. a desired output

    H Y H H W

    H Y H H W E dW d

    T des

    T T

    T des

    T T

    =

    == 0222

    Solving for optimum TX Eq taps, W ( ) desT T ls Y H H H W 1=

    des

    T

    des

    T

    des

    T T Y Y HW Y HW H W E += 22

    Computing the error matrix norm 2

    This will yield a W matrix to produce a value of 1 at the output cursor,i.e. an FIR filter with gain Need to normalize by the total abs(tap) sum for TX FIR realization

    ( ) ( )

    ( )=

    =n

    i ls

    lslsnorm

    nW

    nW nW

    1

  • 8/10/2019 Lecture7 Ee689 Eq Intro Txeq

    23/27

  • 8/10/2019 Lecture7 Ee689 Eq Intro Txeq

    24/27

    TX FIR Circuit Architectures

    Direct FIR vs Segmented DAC Direct FIR Parallel output drivers for output taps Each parallel driver must be sized to

    handle its potential maximum current Lower power & complexity Higher output capacitance

    Segmented DAC Minimum sized output transistors to

    handle peak output current Lowest output capacitance Most power & complexity

    Need mapping table (RAM) Very flexible in equalization

    24

    Segmented DAC

    Direct FIR

    [Zerbe ]

    [Zerbe ]

  • 8/10/2019 Lecture7 Ee689 Eq Intro Txeq

    25/27

    Direct FIR Equalization

    25

    L

    L L

    L

    L

    L

    L

    L

    L

    1x 4x 2x 1x

    1/4 1 1/2 1/4IDACs

    &BiasControl

    sgn -1 sgn 0 sgn 1 sgn 2

    50

    Out-P

    Out-N

    4:2MUX

    2

    2

    2

    21

    D0

    D1

    D2

    D3

    VDDA=1.2VVDD=1.0V

    VDDIO=1.0V

    VDDA=1.2V

    1

    1

    1

    C2 (5GHz)From on-chip PLL

    2

    ( 2

    . 5 G b / s )

    (10Gb/s)

    (5Gb/s)

    ESD

    L

    L L

    L

    L

    L

    L

    L

    L

    LL

    LL LL

    LL

    LL

    LL

    LL

    LL

    LL

    1x 4x 2x 1x

    1/4 1 1/2 1/4IDACs

    &BiasControl

    sgn -1 sgn 0 sgn 1 sgn 2

    50

    Out-P

    Out-N

    4:2MUX

    2

    2

    2

    21

    D0

    D1

    D2

    D3

    VDDA=1.2VVDD=1.0V

    VDDIO=1.0V

    VDDA=1.2V

    1

    1

    1

    C2 (5GHz)From on-chip PLL

    2

    ( 2

    . 5 G b / s )

    (10Gb/s)

    (5Gb/s)

    ESD

    ] 2

    210102101

    TERM out

    R D I D I D I D I V

    A Low Power 10Gb/s Serial Link Transmitter in 90-nmCMOS, A. Rylyakov et al., CSICS 2005

    I -1 I 0 I 1 I 2

    D(1) D(0) D(-1) D(-2)

  • 8/10/2019 Lecture7 Ee689 Eq Intro Txeq

    26/27

    Segmented DAC Example

    26

    [Casper ISSCC 2006 ]Row = 4-bit data patternColumn = 6-bit weighting

    4 filtered bits(parallel) at 6-bit

    resolution

    Sized only todeliver maximum

    total current

  • 8/10/2019 Lecture7 Ee689 Eq Intro Txeq

    27/27

    Next Time

    RX FIR RX CTLE RX DFE

    Alternate/Future Approaches

    27